publication_date
stringlengths 20
20
| doc_url
stringlengths 33
34
| id
stringlengths 33
34
| authors
stringlengths 5
20.1k
| update_date
stringlengths 20
20
| category_all
stringlengths 5
138
| abstract
stringlengths 123
3.09k
| category_primary
stringclasses 78
values | title
stringlengths 10
211
|
|---|---|---|---|---|---|---|---|---|
2022-12-20T18:50:08Z | http://arxiv.org/pdf/2212.10544v2 | http://arxiv.org/abs/2212.10544v2 | Junxiong Wang, Jing Nathan Yan, Albert Gu, Alexander M. Rush | 2023-05-09T01:08:20Z | cs.CL, cs.LG | Transformers have been essential to pretraining success in NLP. While other
architectures have been used, downstream accuracy is either significantly
worse, or requires attention layers to match standard benchmarks such as GLUE.
This work explores pretraining without attention by using recent advances in
sequence routing based on state-space models (SSMs). Our proposed model,
Bidirectional Gated SSM (BiGS), combines SSM layers with a multiplicative
gating architecture that has been effective in simplified sequence modeling
architectures. The model learns static layers that do not consider pair-wise
interactions. Even so, BiGS is able to match BERT pretraining accuracy on GLUE
and can be extended to long-form pretraining of 4096 tokens without
approximation. Analysis shows that while the models have similar average
accuracy, the approach has different inductive biases than BERT in terms of
interactions and syntactic representations. All models from this work are
available at https://github.com/jxiw/BiGS. | cs.CL | Pretraining Without Attention |
2022-12-20T18:50:00Z | http://arxiv.org/pdf/2212.10543v2 | http://arxiv.org/abs/2212.10543v2 | Skyler Hallinan, Alisa Liu, Yejin Choi, Maarten Sap | 2023-05-26T20:26:06Z | cs.CL, cs.AI | Text detoxification has the potential to mitigate the harms of toxicity by
rephrasing text to remove offensive meaning, but subtle toxicity remains
challenging to tackle. We introduce MaRCo, a detoxification algorithm that
combines controllable generation and text rewriting methods using a Product of
Experts with autoencoder language models (LMs). MaRCo uses likelihoods under a
non-toxic LM (expert) and a toxic LM (anti-expert) to find candidate words to
mask and potentially replace. We evaluate our method on several subtle toxicity
and microaggressions datasets, and show that it not only outperforms baselines
on automatic metrics, but MaRCo's rewrites are preferred 2.1 $\times$ more in
human evaluation. Its applicability to instances of subtle toxicity is
especially promising, demonstrating a path forward for addressing increasingly
elusive online hate. | cs.CL | Detoxifying Text with MaRCo: Controllable Revision with Experts and Anti-Experts |
2022-12-20T18:47:13Z | http://arxiv.org/pdf/2212.10539v1 | http://arxiv.org/abs/2212.10539v1 | Weijia Shi, Xiaochuang Han, Hila Gonen, Ari Holtzman, Yulia Tsvetkov, Luke Zettlemoyer | 2022-12-20T18:47:13Z | cs.CL | Large language models can perform new tasks in a zero-shot fashion, given
natural language prompts that specify the desired behavior. Such prompts are
typically hand engineered, but can also be learned with gradient-based methods
from labeled data. However, it is underexplored what factors make the prompts
effective, especially when the prompts are natural language. In this paper, we
investigate common attributes shared by effective prompts. We first propose a
human readable prompt tuning method (F LUENT P ROMPT) based on Langevin
dynamics that incorporates a fluency constraint to find a diverse distribution
of effective and fluent prompts. Our analysis reveals that effective prompts
are topically related to the task domain and calibrate the prior probability of
label words. Based on these findings, we also propose a method for generating
prompts using only unlabeled data, outperforming strong baselines by an average
of 7.0% accuracy across three tasks. | cs.CL | Toward Human Readable Prompt Tuning: Kubrick's The Shining is a good movie, and a good prompt too? |
2022-12-20T18:46:28Z | http://arxiv.org/pdf/2212.10537v2 | http://arxiv.org/abs/2212.10537v2 | Martha Lewis, Nihal V. Nayak, Peilin Yu, Qinan Yu, Jack Merullo, Stephen H. Bach, Ellie Pavlick | 2023-03-29T15:34:23Z | cs.CV, cs.AI, cs.CL | Large-scale neural network models combining text and images have made
incredible progress in recent years. However, it remains an open question to
what extent such models encode compositional representations of the concepts
over which they operate, such as correctly identifying ''red cube'' by
reasoning over the constituents ''red'' and ''cube''. In this work, we focus on
the ability of a large pretrained vision and language model (CLIP) to encode
compositional concepts and to bind variables in a structure-sensitive way
(e.g., differentiating ''cube behind sphere'' from ''sphere behind cube''). In
order to inspect the performance of CLIP, we compare several architectures from
research on compositional distributional semantics models (CDSMs), a line of
research that attempts to implement traditional compositional linguistic
structures within embedding spaces. We find that CLIP can compose concepts in a
single-object setting, but in situations where concept binding is needed,
performance drops dramatically. At the same time, CDSMs also perform poorly,
with best performance at chance level. | cs.CV | Does CLIP Bind Concepts? Probing Compositionality in Large Image Models |
2022-12-20T18:46:20Z | http://arxiv.org/pdf/2212.10536v1 | http://arxiv.org/abs/2212.10536v1 | Yuling Gu | 2022-12-20T18:46:20Z | cs.CL | Transformer-based language models have shown strong performance on an array
of natural language understanding tasks. However, the question of how these
models react to implicit meaning has been largely unexplored. We investigate
this using the complement coercion phenomenon, which involves sentences like
"The student finished the book about sailing" where the action "reading" is
implicit. We compare LMs' surprisal estimates at various critical sentence
regions in sentences with and without implicit meaning. Effects associated with
recovering implicit meaning were found at a critical region other than where
sentences minimally differ. We then use follow-up experiments to factor out
potential confounds, revealing different perspectives that offer a richer and
more accurate picture. | cs.CL | Measure More, Question More: Experimental Studies on Transformer-based Language Models and Complement Coercion |
2022-12-20T18:46:16Z | http://arxiv.org/pdf/2212.10535v2 | http://arxiv.org/abs/2212.10535v2 | Pan Lu, Liang Qiu, Wenhao Yu, Sean Welleck, Kai-Wei Chang | 2023-06-22T01:37:02Z | cs.AI, cs.CL, cs.CV, cs.LG | Mathematical reasoning is a fundamental aspect of human intelligence and is
applicable in various fields, including science, engineering, finance, and
everyday life. The development of artificial intelligence (AI) systems capable
of solving math problems and proving theorems has garnered significant interest
in the fields of machine learning and natural language processing. For example,
mathematics serves as a testbed for aspects of reasoning that are challenging
for powerful deep learning models, driving new algorithmic and modeling
advances. On the other hand, recent advances in large-scale neural language
models have opened up new benchmarks and opportunities to use deep learning for
mathematical reasoning. In this survey paper, we review the key tasks,
datasets, and methods at the intersection of mathematical reasoning and deep
learning over the past decade. We also evaluate existing benchmarks and
methods, and discuss future research directions in this domain. | cs.AI | A Survey of Deep Learning for Mathematical Reasoning |
2022-12-20T18:46:08Z | http://arxiv.org/pdf/2212.10534v3 | http://arxiv.org/abs/2212.10534v3 | Zeming Chen, Qiyue Gao, Antoine Bosselut, Ashish Sabharwal, Kyle Richardson | 2023-06-05T19:16:25Z | cs.CL | Models trained with counterfactually augmented data learn representations of
the causal structure of tasks, enabling robust generalization. However,
high-quality counterfactual data is scarce for most tasks and not easily
generated at scale. When crowdsourced, such data is typically limited in scale
and diversity; when generated using supervised methods, it is computationally
expensive to extend to new counterfactual dimensions. In this work, we
introduce DISCO (DIStilled COunterfactual Data), a new method for automatically
generating high quality counterfactual data at scale. DISCO engineers prompts
to generate phrasal perturbations with a large general language model. Then, a
task-specific teacher model filters these generations to distill high-quality
counterfactual data. While task-agnostic, we apply our pipeline to the task of
natural language inference (NLI) and find that on challenging evaluations such
as the NLI stress test, comparatively smaller student models trained with DISCO
generated counterfactuals are more robust (6% absolute) and generalize better
across distributions (2%) compared to models trained without data augmentation.
Furthermore, DISCO augmented models are 10% more consistent between
counterfactual pairs on three evaluation sets, demonstrating that DISCO
augmentation enables models to more reliably learn causal representations. Our
repository is available at: https://github.com/eric11eca/disco | cs.CL | DISCO: Distilling Counterfactuals with Large Language Models |
2022-12-20T18:45:12Z | http://arxiv.org/pdf/2212.11672v1 | http://arxiv.org/abs/2212.11672v1 | Rishi Bommasani, Percy Liang | 2022-12-20T18:45:12Z | cs.CL | How do we design measures of social bias that we trust? While prior work has
introduced several measures, no measure has gained widespread trust: instead,
mounting evidence argues we should distrust these measures. In this work, we
design bias measures that warrant trust based on the cross-disciplinary theory
of measurement modeling. To combat the frequently fuzzy treatment of social
bias in NLP, we explicitly define social bias, grounded in principles drawn
from social science research. We operationalize our definition by proposing a
general bias measurement framework DivDist, which we use to instantiate 5
concrete bias measures. To validate our measures, we propose a rigorous testing
protocol with 8 testing criteria (e.g. predictive validity: do measures predict
biases in US employment?). Through our testing, we demonstrate considerable
evidence to trust our measures, showing they overcome conceptual, technical,
and empirical deficiencies present in prior measures. | cs.CL | Trustworthy Social Bias Measurement |
2022-12-20T18:45:07Z | http://arxiv.org/pdf/2212.10529v3 | http://arxiv.org/abs/2212.10529v3 | Xingxuan Li, Yutong Li, Lin Qiu, Shafiq Joty, Lidong Bing | 2024-02-29T13:14:37Z | cs.CL, cs.AI, cs.CY | In this work, we designed unbiased prompts to systematically evaluate the
psychological safety of large language models (LLMs). First, we tested five
different LLMs by using two personality tests: Short Dark Triad (SD-3) and Big
Five Inventory (BFI). All models scored higher than the human average on SD-3,
suggesting a relatively darker personality pattern. Despite being instruction
fine-tuned with safety metrics to reduce toxicity, InstructGPT, GPT-3.5, and
GPT-4 still showed dark personality patterns; these models scored higher than
self-supervised GPT-3 on the Machiavellianism and narcissism traits on SD-3.
Then, we evaluated the LLMs in the GPT series by using well-being tests to
study the impact of fine-tuning with more training data. We observed a
continuous increase in the well-being scores of GPT models. Following these
observations, we showed that fine-tuning Llama-2-chat-7B with responses from
BFI using direct preference optimization could effectively reduce the
psychological toxicity of the model. Based on the findings, we recommended the
application of systematic and comprehensive psychological metrics to further
evaluate and improve the safety of LLMs. | cs.CL | Evaluating Psychological Safety of Large Language Models |
2022-12-20T18:44:21Z | http://arxiv.org/pdf/2212.10528v1 | http://arxiv.org/abs/2212.10528v1 | Jing Lu, Keith Hall, Ji Ma, Jianmo Ni | 2022-12-20T18:44:21Z | cs.CL, cs.IR | We present Hybrid Infused Reranking for Passages Retrieval (HYRR), a
framework for training rerankers based on a hybrid of BM25 and neural retrieval
models. Retrievers based on hybrid models have been shown to outperform both
BM25 and neural models alone. Our approach exploits this improved performance
when training a reranker, leading to a robust reranking model. The reranker, a
cross-attention neural model, is shown to be robust to different first-stage
retrieval systems, achieving better performance than rerankers simply trained
upon the first-stage retrievers in the multi-stage systems. We present
evaluations on a supervised passage retrieval task using MS MARCO and zero-shot
retrieval tasks using BEIR. The empirical results show strong performance on
both evaluations. | cs.CL | HYRR: Hybrid Infused Reranking for Passage Retrieval |
2022-12-20T18:41:38Z | http://arxiv.org/pdf/2212.10526v3 | http://arxiv.org/abs/2212.10526v3 | John Giorgi, Luca Soldaini, Bo Wang, Gary Bader, Kyle Lo, Lucy Lu Wang, Arman Cohan | 2023-10-25T13:25:20Z | cs.CL, cs.AI | Multi-document summarization (MDS) assumes a set of topic-related documents
are provided as input. In practice, this document set is not always available;
it would need to be retrieved given an information need, i.e. a question or
topic statement, a setting we dub "open-domain" MDS. We study this more
challenging setting by formalizing the task and bootstrapping it using existing
datasets, retrievers and summarizers. Via extensive automatic and human
evaluation, we determine: (1) state-of-the-art summarizers suffer large
reductions in performance when applied to open-domain MDS, (2) additional
training in the open-domain setting can reduce this sensitivity to imperfect
retrieval, and (3) summarizers are insensitive to the retrieval of duplicate
documents and the order of retrieved documents, but highly sensitive to other
errors, like the retrieval of irrelevant documents. Based on our results, we
provide practical guidelines to enable future work on open-domain MDS, e.g. how
to choose the number of retrieved documents to summarize. Our results suggest
that new retrieval and summarization methods and annotated resources for
training and evaluation are necessary for further progress in the open-domain
setting. | cs.CL | Open Domain Multi-document Summarization: A Comprehensive Study of Model Brittleness under Retrieval |
2022-12-20T18:39:59Z | http://arxiv.org/pdf/2212.10525v2 | http://arxiv.org/abs/2212.10525v2 | Suwon Shon, Siddhant Arora, Chyi-Jiunn Lin, Ankita Pasad, Felix Wu, Roshan Sharma, Wei-Lun Wu, Hung-Yi Lee, Karen Livescu, Shinji Watanabe | 2023-06-15T22:51:09Z | cs.CL, eess.AS | Spoken language understanding (SLU) tasks have been studied for many decades
in the speech research community, but have not received as much attention as
lower-level tasks like speech and speaker recognition. In particular, there are
not nearly as many SLU task benchmarks, and many of the existing ones use data
that is not freely available to all researchers. Recent work has begun to
introduce such benchmark datasets for several tasks. In this work, we introduce
several new annotated SLU benchmark tasks based on freely available speech
data, which complement existing benchmarks and address gaps in the SLU
evaluation landscape. We contribute four tasks: question answering and
summarization involve inference over longer speech sequences; named entity
localization addresses the speech-specific task of locating the targeted
content in the signal; dialog act classification identifies the function of a
given speech utterance. We follow the blueprint of the Spoken Language
Understanding Evaluation (SLUE) benchmark suite. In order to facilitate the
development of SLU models that leverage the success of pre-trained speech
representations, we will be publishing for each task (i) annotations for a
relatively small fine-tuning set, (ii) annotated development and test sets, and
(iii) baseline models for easy reproducibility and comparisons. In this work,
we present the details of data collection and annotation and the performance of
the baseline models. We also perform sensitivity analysis of pipeline models'
performance (speech recognizer + text model) to the speech recognition
accuracy, using more than 20 state-of-the-art speech recognition models. | cs.CL | SLUE Phase-2: A Benchmark Suite of Diverse Spoken Language Understanding Tasks |
2022-12-20T18:37:11Z | http://arxiv.org/pdf/2212.10522v2 | http://arxiv.org/abs/2212.10522v2 | Yanran Chen, Steffen Eger | 2023-12-26T18:05:42Z | cs.CL | We consider the end-to-end abstract-to-title generation problem, exploring
seven recent transformer based models (including ChatGPT) fine-tuned on more
than 30k abstract-title pairs from NLP and machine learning (ML) venues. As an
extension, we also consider the harder problem of generating humorous paper
titles. For the latter, we compile the first large-scale humor annotated
dataset for scientific papers in the NLP/ML domains, comprising almost ~2.6k
titles. We evaluate all models using human and automatic metrics. Our human
evaluation suggests that our best end-to-end system performs similarly to human
authors (but arguably slightly worse). Generating funny titles is more
difficult, however, and our automatic systems clearly underperform relative to
humans and often learn dataset artefacts of humor. Finally, ChatGPT, without
any fine-tuning, performs on the level of our best fine-tuned system. | cs.CL | Transformers Go for the LOLs: Generating (Humourous) Titles from Scientific Abstracts End-to-End |
2022-12-20T18:35:33Z | http://arxiv.org/pdf/2212.10711v1 | http://arxiv.org/abs/2212.10711v1 | Alex Tamkin, Kunal Handa, Avash Shrestha, Noah Goodman | 2022-12-20T18:35:33Z | cs.CL, cs.LG | Language models have recently achieved strong performance across a wide range
of NLP benchmarks. However, unlike benchmarks, real world tasks are often
poorly specified, and agents must deduce the user's intended behavior from a
combination of context, instructions, and examples. We investigate how both
humans and models behave in the face of such task ambiguity by proposing
AmbiBench, a new benchmark of six ambiguously-specified classification tasks.
We evaluate humans and models on AmbiBench by seeing how well they identify the
intended task using 1) instructions with varying degrees of ambiguity, and 2)
different numbers of labeled examples. We find that the combination of model
scaling (to 175B parameters) and training with human feedback data enables
models to approach or exceed the accuracy of human participants across tasks,
but that either one alone is not sufficient. In addition, we show how to
dramatically improve the accuracy of language models trained without
large-scale human feedback training by finetuning on a small number of
ambiguous in-context examples, providing a promising direction for teaching
models to generalize well in the face of ambiguity. | cs.CL | Task Ambiguity in Humans and Language Models |
2022-12-20T18:35:21Z | http://arxiv.org/pdf/2212.10520v3 | http://arxiv.org/abs/2212.10520v3 | Fatemehsadat Mireshghallah, Yu Su, Tatsunori Hashimoto, Jason Eisner, Richard Shin | 2023-06-08T18:03:43Z | cs.CL | Task-oriented dialogue systems often assist users with personal or
confidential matters. For this reason, the developers of such a system are
generally prohibited from observing actual usage. So how can they know where
the system is failing and needs more training data or new functionality? In
this work, we study ways in which realistic user utterances can be generated
synthetically, to help increase the linguistic and functional coverage of the
system, without compromising the privacy of actual users. To this end, we
propose a two-stage Differentially Private (DP) generation method which first
generates latent semantic parses, and then generates utterances based on the
parses. Our proposed approach improves MAUVE by 2.5$\times$ and parse tree
function type overlap by 1.3$\times$ relative to current approaches for private
synthetic data generation, improving both on fluency and semantic coverage. We
further validate our approach on a realistic domain adaptation task of adding
new functionality from private user data to a semantic parser, and show overall
gains of 8.5% points in accuracy with the new feature. | cs.CL | Privacy-Preserving Domain Adaptation of Semantic Parsers |
2022-12-20T18:31:50Z | http://arxiv.org/pdf/2212.10515v2 | http://arxiv.org/abs/2212.10515v2 | Yi-Lin Tuan, Alon Albalak, Wenda Xu, Michael Saxon, Connor Pryor, Lise Getoor, William Yang Wang | 2023-07-08T21:59:43Z | cs.CL | Despite their widespread adoption, neural conversation models have yet to
exhibit natural chat capabilities with humans. In this research, we examine
user utterances as causes and generated responses as effects, recognizing that
changes in a cause should produce a different effect. To further explore this
concept, we have compiled and expanded upon a new dataset called CausalDialogue
through crowd-sourcing. This dataset includes multiple cause-effect pairs
within a directed acyclic graph (DAG) structure. Our analysis reveals that
traditional loss functions struggle to effectively incorporate the DAG
structure, leading us to propose a causality-enhanced method called Exponential
Maximum Average Treatment Effect (ExMATE) to enhance the impact of causality at
the utterance level in training neural conversation models. To evaluate the
needs of considering causality in dialogue generation, we built a comprehensive
benchmark on CausalDialogue dataset using different models, inference, and
training methods. Through experiments, we find that a causality-inspired loss
like ExMATE can improve the diversity and agility of conventional loss function
and there is still room for improvement to reach human-level quality on this
new dataset. | cs.CL | CausalDialogue: Modeling Utterance-level Causality in Conversations |
2022-12-20T18:30:15Z | http://arxiv.org/pdf/2212.10511v4 | http://arxiv.org/abs/2212.10511v4 | Alex Mallen, Akari Asai, Victor Zhong, Rajarshi Das, Daniel Khashabi, Hannaneh Hajishirzi | 2023-07-02T07:21:59Z | cs.CL, cs.AI, cs.LG | Despite their impressive performance on diverse tasks, large language models
(LMs) still struggle with tasks requiring rich world knowledge, implying the
limitations of relying solely on their parameters to encode a wealth of world
knowledge. This paper aims to understand LMs' strengths and limitations in
memorizing factual knowledge, by conducting large-scale knowledge probing
experiments of 10 models and 4 augmentation methods on PopQA, our new
open-domain QA dataset with 14k questions. We find that LMs struggle with less
popular factual knowledge, and that scaling fails to appreciably improve
memorization of factual knowledge in the long tail. We then show that
retrieval-augmented LMs largely outperform orders of magnitude larger LMs,
while unassisted LMs remain competitive in questions about high-popularity
entities. Based on those findings, we devise a simple, yet effective, method
for powerful and efficient retrieval-augmented LMs, which retrieves
non-parametric memories only when necessary. Experimental results show that
this significantly improves models' performance while reducing the inference
costs. | cs.CL | When Not to Trust Language Models: Investigating Effectiveness of Parametric and Non-Parametric Memories |
2022-12-20T18:26:34Z | http://arxiv.org/pdf/2212.10509v2 | http://arxiv.org/abs/2212.10509v2 | Harsh Trivedi, Niranjan Balasubramanian, Tushar Khot, Ashish Sabharwal | 2023-06-23T00:59:13Z | cs.CL | Prompting-based large language models (LLMs) are surprisingly powerful at
generating natural language reasoning steps or Chains-of-Thoughts (CoT) for
multi-step question answering (QA). They struggle, however, when the necessary
knowledge is either unavailable to the LLM or not up-to-date within its
parameters. While using the question to retrieve relevant text from an external
knowledge source helps LLMs, we observe that this one-step retrieve-and-read
approach is insufficient for multi-step QA. Here, \textit{what to retrieve}
depends on \textit{what has already been derived}, which in turn may depend on
\textit{what was previously retrieved}. To address this, we propose IRCoT, a
new approach for multi-step QA that interleaves retrieval with steps
(sentences) in a CoT, guiding the retrieval with CoT and in turn using
retrieved results to improve CoT. Using IRCoT with GPT3 substantially improves
retrieval (up to 21 points) as well as downstream QA (up to 15 points) on four
datasets: HotpotQA, 2WikiMultihopQA, MuSiQue, and IIRC. We observe similar
substantial gains in out-of-distribution (OOD) settings as well as with much
smaller models such as Flan-T5-large without additional training. IRCoT reduces
model hallucination, resulting in factually more accurate CoT reasoning. Code,
data, and prompts are available at \url{https://github.com/stonybrooknlp/ircot} | cs.CL | Interleaving Retrieval with Chain-of-Thought Reasoning for Knowledge-Intensive Multi-Step Questions |
2022-12-20T18:20:50Z | http://arxiv.org/pdf/2212.10505v2 | http://arxiv.org/abs/2212.10505v2 | Fangyu Liu, Julian Martin Eisenschlos, Francesco Piccinno, Syrine Krichene, Chenxi Pang, Kenton Lee, Mandar Joshi, Wenhu Chen, Nigel Collier, Yasemin Altun | 2023-05-23T18:28:39Z | cs.CL, cs.AI, cs.CV | Visual language such as charts and plots is ubiquitous in the human world.
Comprehending plots and charts requires strong reasoning skills. Prior
state-of-the-art (SOTA) models require at least tens of thousands of training
examples and their reasoning capabilities are still much limited, especially on
complex human-written queries. This paper presents the first one-shot solution
to visual language reasoning. We decompose the challenge of visual language
reasoning into two steps: (1) plot-to-text translation, and (2) reasoning over
the translated text. The key in this method is a modality conversion module,
named as DePlot, which translates the image of a plot or chart to a linearized
table. The output of DePlot can then be directly used to prompt a pretrained
large language model (LLM), exploiting the few-shot reasoning capabilities of
LLMs. To obtain DePlot, we standardize the plot-to-table task by establishing
unified task formats and metrics, and train DePlot end-to-end on this task.
DePlot can then be used off-the-shelf together with LLMs in a plug-and-play
fashion. Compared with a SOTA model finetuned on more than >28k data points,
DePlot+LLM with just one-shot prompting achieves a 24.0% improvement over
finetuned SOTA on human-written queries from the task of chart QA. | cs.CL | DePlot: One-shot visual language reasoning by plot-to-table translation |
2022-12-20T18:18:41Z | http://arxiv.org/pdf/2212.10504v2 | http://arxiv.org/abs/2212.10504v2 | Sang-Woo Lee, Sungdong Kim, Donghyeon Ko, Donghoon Ham, Youngki Hong, Shin Ah Oh, Hyunhoon Jung, Wangkyo Jung, Kyunghyun Cho, Donghyun Kwak, Hyungsuk Noh, Woomyoung Park | 2023-05-24T11:46:33Z | cs.CL | Task-oriented dialogue (TOD) systems are mainly based on the
slot-filling-based TOD (SF-TOD) framework, in which dialogues are broken down
into smaller, controllable units (i.e., slots) to fulfill a specific task. A
series of approaches based on this framework achieved remarkable success on
various TOD benchmarks. However, we argue that the current TOD benchmarks are
limited to surrogate real-world scenarios and that the current TOD models are
still a long way to cover the scenarios. In this position paper, we first
identify current status and limitations of SF-TOD systems. After that, we
explore the WebTOD framework, the alternative direction for building a scalable
TOD system when a web/mobile interface is available. In WebTOD, the dialogue
system learns how to understand the web/mobile interface that the human agent
interacts with, powered by a large-scale language model. | cs.CL | Can Current Task-oriented Dialogue Models Automate Real-world Scenarios in the Wild? |
2022-12-20T18:17:28Z | http://arxiv.org/pdf/2212.10503v2 | http://arxiv.org/abs/2212.10503v2 | Kelly Marchisio, Patrick Lewis, Yihong Chen, Mikel Artetxe | 2023-07-04T19:06:55Z | cs.CL, cs.LG | Prior work shows that it is possible to expand pretrained Masked Language
Models (MLMs) to new languages by learning a new set of embeddings, while
keeping the transformer body frozen. Despite learning a small subset of
parameters, this approach is not compute-efficient, as training the new
embeddings requires a full forward and backward pass over the entire model. We
propose mini-model adaptation, a compute-efficient alternative that builds a
shallow mini-model from a fraction of a large model's parameters. New
language-specific embeddings can then be efficiently trained over the
mini-model and plugged into the aligned large model for rapid cross-lingual
transfer. We explore two approaches to learn mini-models: MiniJoint, which
jointly pretrains the primary model and the mini-model using a single
transformer with a secondary MLM head at a middle layer; and MiniPost, where we
start from a regular pretrained model, build a mini-model by extracting and
freezing a few layers, and learn a small number of parameters on top.
Experiments on XNLI, MLQA and PAWS-X show that mini-model adaptation matches
the performance of the standard approach using 2.3x less compute on average. | cs.CL | Mini-Model Adaptation: Efficiently Extending Pretrained Models to New Languages via Aligned Shallow Training |
2022-12-20T18:17:11Z | http://arxiv.org/pdf/2212.10502v2 | http://arxiv.org/abs/2212.10502v2 | Li Du, Lucas Torroba Hennigen, Tiago Pimentel, Clara Meister, Jason Eisner, Ryan Cotterell | 2023-08-21T18:01:57Z | cs.CL, cs.FL | Language modeling, a central task in natural language processing, involves
estimating a probability distribution over strings. In most cases, the
estimated distribution sums to 1 over all finite strings. However, in some
pathological cases, probability mass can ``leak'' onto the set of infinite
sequences. In order to characterize the notion of leakage more precisely, this
paper offers a measure-theoretic treatment of language modeling. We prove that
many popular language model families are in fact tight, meaning that they will
not leak in this sense. We also generalize characterizations of tightness
proposed in previous works. | cs.CL | A Measure-Theoretic Characterization of Tight Language Models |
2022-12-20T18:12:49Z | http://arxiv.org/pdf/2212.10498v2 | http://arxiv.org/abs/2212.10498v2 | Elron Bandel, Yoav Katz, Noam Slonim, Liat Ein-Dor | 2022-12-22T16:26:36Z | cs.CL | Attribute-controlled text rewriting, also known as text style-transfer, has a
crucial role in regulating attributes and biases of textual training data and a
machine generated text. In this work we present SimpleStyle, a minimalist yet
effective approach for style-transfer composed of two simple ingredients:
controlled denoising and output filtering. Despite the simplicity of our
approach, which can be succinctly described with a few lines of code, it is
competitive with previous state-of-the-art methods both in automatic and in
human evaluation. To demonstrate the adaptability and practical value of our
system beyond academic data, we apply SimpleStyle to transfer a wide range of
text attributes appearing in real-world textual data from social networks.
Additionally, we introduce a novel "soft noising" technique that further
improves the performance of our system. We also show that teaching a student
model to generate the output of SimpleStyle can result in a system that
performs style transfer of equivalent quality with only a single greedy-decoded
sample. Finally, we suggest our method as a remedy for the fundamental
incompatible baseline issue that holds progress in the field. We offer our
protocol as a simple yet strong baseline for works that wish to make
incremental advancements in the field of attribute controlled text rewriting. | cs.CL | SimpleStyle: An Adaptable Style Transfer Approach |
2022-12-20T18:09:52Z | http://arxiv.org/pdf/2212.10496v1 | http://arxiv.org/abs/2212.10496v1 | Luyu Gao, Xueguang Ma, Jimmy Lin, Jamie Callan | 2022-12-20T18:09:52Z | cs.IR, cs.CL | While dense retrieval has been shown effective and efficient across tasks and
languages, it remains difficult to create effective fully zero-shot dense
retrieval systems when no relevance label is available. In this paper, we
recognize the difficulty of zero-shot learning and encoding relevance. Instead,
we propose to pivot through Hypothetical Document Embeddings~(HyDE). Given a
query, HyDE first zero-shot instructs an instruction-following language model
(e.g. InstructGPT) to generate a hypothetical document. The document captures
relevance patterns but is unreal and may contain false details. Then, an
unsupervised contrastively learned encoder~(e.g. Contriever) encodes the
document into an embedding vector. This vector identifies a neighborhood in the
corpus embedding space, where similar real documents are retrieved based on
vector similarity. This second step ground the generated document to the actual
corpus, with the encoder's dense bottleneck filtering out the incorrect
details. Our experiments show that HyDE significantly outperforms the
state-of-the-art unsupervised dense retriever Contriever and shows strong
performance comparable to fine-tuned retrievers, across various tasks (e.g. web
search, QA, fact verification) and languages~(e.g. sw, ko, ja). | cs.IR | Precise Zero-Shot Dense Retrieval without Relevance Labels |
2022-12-20T17:54:37Z | http://arxiv.org/pdf/2212.10481v2 | http://arxiv.org/abs/2212.10481v2 | Zhiruo Wang, Shuyan Zhou, Daniel Fried, Graham Neubig | 2023-05-19T14:27:46Z | cs.SE, cs.AI, cs.CL | To extend the scope of coding queries to more realistic settings, we propose
ODEX, the first Open-Domain EXecution-based natural language (NL) to Python
code generation dataset. ODEX has 945 NL-Code pairs spanning 79 diverse
libraries, along with 1,707 human-written test cases for execution. Our NL-Code
pairs are harvested from StackOverflow forums to encourage natural and
practical coding queries. Moreover, ODEX supports four natural languages as
intents, in English, Spanish, Japanese, and Russian. ODEX unveils intriguing
behavioral differences among top-performing code language models (LM). While
CODEX achieves better overall results, CODEGEN improves effectively via scaling
-- CODEGEN 6.1B performs comparably with CODEX 12B. Both models show
substantial gaps between open and closed domains, but CODEGEN gaps tend to
decrease with model size while CODEX gaps increase. We release ODEX to
facilitate research into open-domain problems for the code generation
community. | cs.SE | Execution-Based Evaluation for Open-Domain Code Generation |
2022-12-20T17:54:08Z | http://arxiv.org/pdf/2212.11119v1 | http://arxiv.org/abs/2212.11119v1 | Gustavo Henrique de Rosa, João Paulo Papa | 2022-12-20T17:54:08Z | cs.CL, cs.AI, cs.LG | This work presents a thorough review concerning recent studies and text
generation advancements using Generative Adversarial Networks. The usage of
adversarial learning for text generation is promising as it provides
alternatives to generate the so-called "natural" language. Nevertheless,
adversarial text generation is not a simple task as its foremost architecture,
the Generative Adversarial Networks, were designed to cope with continuous
information (image) instead of discrete data (text). Thus, most works are based
on three possible options, i.e., Gumbel-Softmax differentiation, Reinforcement
Learning, and modified training objectives. All alternatives are reviewed in
this survey as they present the most recent approaches for generating text
using adversarial-based techniques. The selected works were taken from renowned
databases, such as Science Direct, IEEEXplore, Springer, Association for
Computing Machinery, and arXiv, whereas each selected work has been critically
analyzed and assessed to present its objective, methodology, and experimental
results. | cs.CL | A survey on text generation using generative adversarial networks |
2022-12-20T17:49:49Z | http://arxiv.org/pdf/2212.10474v2 | http://arxiv.org/abs/2212.10474v2 | Jonas Belouadi, Steffen Eger | 2023-05-22T21:15:06Z | cs.CL | State-of-the-art poetry generation systems are often complex. They either
consist of task-specific model pipelines, incorporate prior knowledge in the
form of manually created constraints, or both. In contrast, end-to-end models
would not suffer from the overhead of having to model prior knowledge and could
learn the nuances of poetry from data alone, reducing the degree of human
supervision required. In this work, we investigate end-to-end poetry generation
conditioned on styles such as rhyme, meter, and alliteration. We identify and
address lack of training data and mismatching tokenization algorithms as
possible limitations of past attempts. In particular, we successfully pre-train
ByGPT5, a new token-free decoder-only language model, and fine-tune it on a
large custom corpus of English and German quatrains annotated with our styles.
We show that ByGPT5 outperforms other models such as mT5, ByT5, GPT-2 and
ChatGPT, while also being more parameter efficient and performing favorably
compared to humans. In addition, we analyze its runtime performance and
demonstrate that it is not prone to memorization. We make our code, models, and
datasets publicly available. | cs.CL | ByGPT5: End-to-End Style-conditioned Poetry Generation with Token-free Language Models |
2022-12-20T17:49:27Z | http://arxiv.org/pdf/2212.11670v1 | http://arxiv.org/abs/2212.11670v1 | Rishi Bommasani | 2022-12-20T17:49:27Z | cs.CL, cs.AI, cs.LG | Evaluation is the central means for assessing, understanding, and
communicating about NLP models. In this position paper, we argue evaluation
should be more than that: it is a force for driving change, carrying a
sociological and political character beyond its technical dimensions. As a
force, evaluation's power arises from its adoption: under our view, evaluation
succeeds when it achieves the desired change in the field. Further, by framing
evaluation as a force, we consider how it competes with other forces. Under our
analysis, we conjecture that the current trajectory of NLP suggests
evaluation's power is waning, in spite of its potential for realizing more
pluralistic ambitions in the field. We conclude by discussing the legitimacy of
this power, who acquires this power and how it distributes. Ultimately, we hope
the research community will more aggressively harness evaluation for change. | cs.CL | Evaluation for Change |
2022-12-20T17:42:16Z | http://arxiv.org/pdf/2212.10471v3 | http://arxiv.org/abs/2212.10471v3 | Evgeniia Razumovskaia, Joshua Maynez, Annie Louis, Mirella Lapata, Shashi Narayan | 2024-03-25T17:54:21Z | cs.CL | Previous work has demonstrated the effectiveness of planning for story
generation exclusively in a monolingual setting focusing primarily on English.
We consider whether planning brings advantages to automatic story generation
across languages. We propose a new task of cross-lingual story generation with
planning and present a new dataset for this task. We conduct a comprehensive
study of different plans and generate stories in several languages, by
leveraging the creative and reasoning capabilities of large pre-trained
language models. Our results demonstrate that plans which structure stories
into three acts lead to more coherent and interesting narratives, while
allowing to explicitly control their content and structure. | cs.CL | Little Red Riding Hood Goes Around the Globe:Crosslingual Story Planning and Generation with Large Language Models |
2022-12-20T17:41:18Z | http://arxiv.org/pdf/2212.10469v2 | http://arxiv.org/abs/2212.10469v2 | Christoph Leiter, Hoa Nguyen, Steffen Eger | 2024-02-17T16:55:27Z | cs.CL | State-of-the-art natural language generation evaluation metrics are based on
black-box language models. Hence, recent works consider their explainability
with the goals of better understandability for humans and better metric
analysis, including failure cases. In contrast, our proposed method BMX:
Boosting Natural Language Generation Metrics with explainability explicitly
leverages explanations to boost the metrics' performance. In particular, we
perceive feature importance explanations as word-level scores, which we
convert, via power means, into a segment-level score. We then combine this
segment-level score with the original metric to obtain a better metric. Our
tests show improvements for multiple metrics across MT and summarization
datasets. While improvements in machine translation are small, they are strong
for summarization. Notably, BMX with the LIME explainer and preselected
parameters achieves an average improvement of 0.087 points in Spearman
correlation on the system-level evaluation of SummEval. | cs.CL | BMX: Boosting Natural Language Generation Metrics with Explainability |
2022-12-20T17:40:03Z | http://arxiv.org/pdf/2212.10467v2 | http://arxiv.org/abs/2212.10467v2 | Yu Feng, Ben Zhou, Haoyu Wang, Helen Jin, Dan Roth | 2023-05-31T17:54:08Z | cs.CL | Temporal reasoning is the task of predicting temporal relations of event
pairs. While temporal reasoning models can perform reasonably well on in-domain
benchmarks, we have little idea of these systems' generalizability due to
existing datasets' limitations. In this work, we introduce a novel task named
TODAY that bridges this gap with temporal differential analysis, which as the
name suggests, evaluates whether systems can correctly understand the effect of
incremental changes. Specifically, TODAY introduces slight contextual changes
for given event pairs, and systems are asked to tell how this subtle contextual
change would affect relevant temporal relation distributions. To facilitate
learning, TODAY also annotates human explanations. We show that existing
models, including GPT-3.5, drop to random guessing on TODAY, suggesting that
they heavily rely on spurious information rather than proper reasoning for
temporal predictions. On the other hand, we show that TODAY's supervision style
and explanation annotations can be used in joint learning, encouraging models
to use more appropriate signals during training and thus outperform across
several benchmarks. TODAY can also be used to train models to solicit
incidental supervision from noisy sources such as GPT-3.5, thus moving us more
toward the goal of generic temporal reasoning systems. | cs.CL | Generic Temporal Reasoning with Differential Analysis and Explanation |
2022-12-20T17:39:21Z | http://arxiv.org/pdf/2212.10466v1 | http://arxiv.org/abs/2212.10466v1 | Howard Chen, Huihan Li, Danqi Chen, Karthik Narasimhan | 2022-12-20T17:39:21Z | cs.CL | We consider the task of text generation in language models with constraints
specified in natural language. To this end, we first create a challenging
benchmark Cognac that provides as input to the model a topic with example text,
along with a constraint on text to be avoided. Unlike prior work, our benchmark
contains knowledge-intensive constraints sourced from databases like Wordnet
and Wikidata, which allows for straightforward evaluation while striking a
balance between broad attribute-level and narrow lexical-level controls. We
find that even state-of-the-art language models like GPT-3 fail often on this
task, and propose a solution to leverage a language model's own internal
knowledge to guide generation. Our method, called CognacGen, first queries the
language model to generate guidance terms for a specified topic or constraint,
and uses the guidance to modify the model's token generation probabilities. We
propose three forms of guidance (binary verifier, top-k tokens, textual
example), and employ prefix-tuning approaches to distill the guidance to tackle
diverse natural language constraints. Through extensive empirical evaluations,
we demonstrate that CognacGen can successfully generalize to unseen
instructions and outperform competitive baselines in generating constraint
conforming text. | cs.CL | Controllable Text Generation with Language Constraints |
2022-12-20T17:38:47Z | http://arxiv.org/pdf/2212.10465v3 | http://arxiv.org/abs/2212.10465v3 | Hyunwoo Kim, Jack Hessel, Liwei Jiang, Peter West, Ximing Lu, Youngjae Yu, Pei Zhou, Ronan Le Bras, Malihe Alikhani, Gunhee Kim, Maarten Sap, Yejin Choi | 2023-10-23T18:46:21Z | cs.CL | Data scarcity has been a long standing issue in the field of open-domain
social dialogue. To quench this thirst, we present SODA: the first publicly
available, million-scale high-quality social dialogue dataset. By
contextualizing social commonsense knowledge from a knowledge graph, we are
able to distill an exceptionally broad spectrum of social interactions from a
large language model. Human evaluation shows that conversations in SODA are
more consistent, specific, and (surprisingly) natural than those in prior
human-authored datasets.
Using SODA, we train COSMO: a generalizable conversation model that is
significantly more natural and consistent on unseen datasets than
best-performing conversation models (e.g., GODEL, BlenderBot-1, Koala, Vicuna).
Experiments reveal COSMO is sometimes even preferred to the original
human-written gold responses. Additionally, our results shed light on the
distinction between knowledge-enriched conversations and natural social
chitchats. We plan to make our data, model, and code public. | cs.CL | SODA: Million-scale Dialogue Distillation with Social Commonsense Contextualization |
2022-12-20T17:36:49Z | http://arxiv.org/pdf/2212.10461v1 | http://arxiv.org/abs/2212.10461v1 | Jingjing Xu, Qingxiu Dong, Hongyi Liu, Lei Li | 2022-12-20T17:36:49Z | cs.CL | With increasing scale, large language models demonstrate both quantitative
improvement and new qualitative capabilities, especially as zero-shot learners,
like GPT-3. However, these results rely heavily on delicate prompt design and
large computation. In this work, we explore whether the strong zero-shot
ability could be achieved at a smaller model scale without any external
supervised data. To achieve this goal, we revisit masked language modeling and
present a geometry-guided self-supervised learning method (Go-tuningfor short)
by taking a small number of task-aware self-supervised data to update language
models further. Experiments show that Go-tuning can enable T5-small (80M)
competitive zero-shot results compared with large language models, such as
T5-XL (3B). We also apply Go-tuning on multi-task settings and develop a
multi-task model, mgo-T5 (250M). It can reach the average performance of OPT
(175B) on 9 datasets. | cs.CL | Go-tuning: Improving Zero-shot Learning Abilities of Smaller Language Models |
2022-12-20T17:34:25Z | http://arxiv.org/pdf/2212.10455v2 | http://arxiv.org/abs/2212.10455v2 | Nikita Moghe, Evgeniia Razumovskaia, Liane Guillou, Ivan Vulić, Anna Korhonen, Alexandra Birch | 2023-06-19T04:09:37Z | cs.CL | Task-oriented dialogue (TOD) systems have been widely deployed in many
industries as they deliver more efficient customer support. These systems are
typically constructed for a single domain or language and do not generalise
well beyond this. To support work on Natural Language Understanding (NLU) in
TOD across multiple languages and domains simultaneously, we constructed
MULTI3NLU++, a multilingual, multi-intent, multi-domain dataset. MULTI3NLU++
extends the English only NLU++ dataset to include manual translations into a
range of high, medium, and low resource languages (Spanish, Marathi, Turkish
and Amharic), in two domains (BANKING and HOTELS). Because of its multi-intent
property, MULTI3NLU++ represents complex and natural user goals, and therefore
allows us to measure the realistic performance of TOD systems in a varied set
of the world's languages. We use MULTI3NLU++ to benchmark state-of-the-art
multilingual models for the NLU tasks of intent detection and slot labelling
for TOD systems in the multilingual setting. The results demonstrate the
challenging nature of the dataset, particularly in the low-resource language
setting, offering ample room for future experimentation in multi-domain
multilingual TOD setups. | cs.CL | MULTI3NLU++: A Multilingual, Multi-Intent, Multi-Domain Dataset for Natural Language Understanding in Task-Oriented Dialogue |
2022-12-20T17:28:41Z | http://arxiv.org/pdf/2212.10450v2 | http://arxiv.org/abs/2212.10450v2 | Bosheng Ding, Chengwei Qin, Linlin Liu, Yew Ken Chia, Shafiq Joty, Boyang Li, Lidong Bing | 2023-06-14T16:11:50Z | cs.CL | Data annotation is the process of labeling data that could be used to train
machine learning models. Having high-quality annotation is crucial, as it
allows the model to learn the relationship between the input data and the
desired output. GPT-3, a large-scale language model developed by OpenAI, has
demonstrated impressive zero- and few-shot performance on a wide range of NLP
tasks. It is therefore natural to wonder whether it can be used to effectively
annotate data for NLP tasks. In this paper, we evaluate the performance of
GPT-3 as a data annotator by comparing it with traditional data annotation
methods and analyzing its output on a range of tasks. Through this analysis, we
aim to provide insight into the potential of GPT-3 as a general-purpose data
annotator in NLP. | cs.CL | Is GPT-3 a Good Data Annotator? |
2022-12-20T17:27:10Z | http://arxiv.org/pdf/2212.10449v3 | http://arxiv.org/abs/2212.10449v3 | Artidoro Pagnoni, Alexander R. Fabbri, Wojciech Kryściński, Chien-Sheng Wu | 2023-06-08T22:43:58Z | cs.CL | In long document controllable summarization, where labeled data is scarce,
pretrained models struggle to adapt to the task and effectively respond to user
queries. In this paper, we introduce Socratic pretraining, a question-driven,
unsupervised pretraining objective specifically designed to improve
controllability in summarization tasks. By training a model to generate and
answer relevant questions in a given context, Socratic pretraining enables the
model to more effectively adhere to user-provided queries and identify relevant
content to be summarized. We demonstrate the effectiveness of this approach
through extensive experimentation on two summarization domains, short stories
and dialogue, and multiple control strategies: keywords, questions, and factoid
QA pairs. Our pretraining method relies only on unlabeled documents and a
question generation system and outperforms pre-finetuning approaches that use
additional supervised data. Furthermore, our results show that Socratic
pretraining cuts task-specific labeled data requirements in half, is more
faithful to user-provided queries, and achieves state-of-the-art performance on
QMSum and SQuALITY. | cs.CL | Socratic Pretraining: Question-Driven Pretraining for Controllable Summarization |
2022-12-20T17:25:04Z | http://arxiv.org/pdf/2212.10448v1 | http://arxiv.org/abs/2212.10448v1 | Eugene Yang, Suraj Nair, Dawn Lawrie, James Mayfield, Douglas W. Oard | 2022-12-20T17:25:04Z | cs.IR, cs.CL | A popular approach to creating a zero-shot cross-language retrieval model is
to substitute a monolingual pretrained language model in the retrieval model
with a multilingual pretrained language model such as Multilingual BERT. This
multilingual model is fined-tuned to the retrieval task with monolingual data
such as English MS MARCO using the same training recipe as the monolingual
retrieval model used. However, such transferred models suffer from mismatches
in the languages of the input text during training and inference. In this work,
we propose transferring monolingual retrieval models using adapters, a
parameter-efficient component for a transformer network. By adding adapters
pretrained on language tasks for a specific language with task-specific
adapters, prior work has shown that the adapter-enhanced models perform better
than fine-tuning the entire model when transferring across languages in various
NLP tasks. By constructing dense retrieval models with adapters, we show that
models trained with monolingual data are more effective than fine-tuning the
entire model when transferring to a Cross Language Information Retrieval (CLIR)
setting. However, we found that the prior suggestion of replacing the language
adapters to match the target language at inference time is suboptimal for dense
retrieval models. We provide an in-depth analysis of this discrepancy between
other cross-language NLP tasks and CLIR. | cs.IR | Parameter-efficient Zero-shot Transfer for Cross-Language Dense Retrieval with Adapters |
2022-12-20T17:14:45Z | http://arxiv.org/pdf/2212.10440v1 | http://arxiv.org/abs/2212.10440v1 | Tim Jansen, Yangling Tong, Victoria Zevallos, Pedro Ortiz Suarez | 2022-12-20T17:14:45Z | cs.CL | As demand for large corpora increases with the size of current
state-of-the-art language models, using web data as the main part of the
pre-training corpus for these models has become a ubiquitous practice. This, in
turn, has introduced an important challenge for NLP practitioners, as they are
now confronted with the task of developing highly optimized models and
pipelines for pre-processing large quantities of textual data, which implies,
effectively classifying and filtering multilingual, heterogeneous and noisy
data, at web scale. One of the main components of this pre-processing step for
the pre-training corpora of large language models, is the removal of adult and
harmful content. In this paper we explore different methods for detecting adult
and harmful of content in multilingual heterogeneous web data. We first show
how traditional methods in harmful content detection, that seemingly perform
quite well in small and specialized datasets quickly break down when confronted
with heterogeneous noisy web data. We then resort to using a perplexity based
approach but with a twist: Instead of using a so-called "clean" corpus to train
a small language model and then use perplexity so select the documents with low
perplexity, i.e., the documents that resemble this so-called "clean" corpus the
most. We train solely with adult and harmful textual data, and then select the
documents having a perplexity value above a given threshold. This approach will
virtually cluster our documents into two distinct groups, which will greatly
facilitate the choice of the threshold for the perplexity and will also allow
us to obtain higher precision than with the traditional classification methods
for detecting adult and harmful content. | cs.CL | Perplexed by Quality: A Perplexity-based Method for Adult and Harmful Content Detection in Multilingual Heterogeneous Web Data |
2022-12-20T17:00:36Z | http://arxiv.org/pdf/2212.10423v1 | http://arxiv.org/abs/2212.10423v1 | Yucheng Zhou, Tao Shen, Xiubo Geng, Chongyang Tao, Guodong Long, Can Xu, Daxin Jiang | 2022-12-20T17:00:36Z | cs.IR, cs.CL | Long document retrieval aims to fetch query-relevant documents from a
large-scale collection, where knowledge distillation has become de facto to
improve a retriever by mimicking a heterogeneous yet powerful cross-encoder.
However, in contrast to passages or sentences, retrieval on long documents
suffers from the scope hypothesis that a long document may cover multiple
topics. This maximizes their structure heterogeneity and poses a
granular-mismatch issue, leading to an inferior distillation efficacy. In this
work, we propose a new learning framework, fine-grained distillation (FGD), for
long-document retrievers. While preserving the conventional dense retrieval
paradigm, it first produces global-consistent representations crossing
different fine granularity and then applies multi-granular aligned distillation
merely during training. In experiments, we evaluate our framework on two
long-document retrieval benchmarks, which show state-of-the-art performance. | cs.IR | Fine-Grained Distillation for Long Document Retrieval |
2022-12-20T16:59:56Z | http://arxiv.org/pdf/2212.10422v3 | http://arxiv.org/abs/2212.10422v3 | Tommaso Mario Buonocore, Claudio Crema, Alberto Redolfi, Riccardo Bellazzi, Enea Parimbelli | 2023-06-28T08:36:20Z | cs.CL, cs.AI, cs.LG, I.2.7; J.3 | In the era of digital healthcare, the huge volumes of textual information
generated every day in hospitals constitute an essential but underused asset
that could be exploited with task-specific, fine-tuned biomedical language
representation models, improving patient care and management. For such
specialized domains, previous research has shown that fine-tuning models
stemming from broad-coverage checkpoints can largely benefit additional
training rounds over large-scale in-domain resources. However, these resources
are often unreachable for less-resourced languages like Italian, preventing
local medical institutions to employ in-domain adaptation. In order to reduce
this gap, our work investigates two accessible approaches to derive biomedical
language models in languages other than English, taking Italian as a concrete
use-case: one based on neural machine translation of English resources,
favoring quantity over quality; the other based on a high-grade, narrow-scoped
corpus natively written in Italian, thus preferring quality over quantity. Our
study shows that data quantity is a harder constraint than data quality for
biomedical adaptation, but the concatenation of high-quality data can improve
model performance even when dealing with relatively size-limited corpora. The
models published from our investigations have the potential to unlock important
research opportunities for Italian hospitals and academia. Finally, the set of
lessons learned from the study constitutes valuable insights towards a solution
to build biomedical language models that are generalizable to other
less-resourced languages and different domain settings. | cs.CL | Localising In-Domain Adaptation of Transformer-Based Biomedical Language Models |
2022-12-20T16:33:09Z | http://arxiv.org/pdf/2212.10409v3 | http://arxiv.org/abs/2212.10409v3 | Valentina Pyatkin, Jena D. Hwang, Vivek Srikumar, Ximing Lu, Liwei Jiang, Yejin Choi, Chandra Bhagavatula | 2023-05-30T18:59:41Z | cs.CL | Context is everything, even in commonsense moral reasoning. Changing contexts
can flip the moral judgment of an action; "Lying to a friend" is wrong in
general, but may be morally acceptable if it is intended to protect their life.
We present ClarifyDelphi, an interactive system that learns to ask
clarification questions (e.g., why did you lie to your friend?) in order to
elicit additional salient contexts of a social or moral situation. We posit
that questions whose potential answers lead to diverging moral judgments are
the most informative. Thus, we propose a reinforcement learning framework with
a defeasibility reward that aims to maximize the divergence between moral
judgments of hypothetical answers to a question. Human evaluation demonstrates
that our system generates more relevant, informative and defeasible questions
compared to competitive baselines. Our work is ultimately inspired by studies
in cognitive science that have investigated the flexibility in moral cognition
(i.e., the diverse contexts in which moral rules can be bent), and we hope that
research in this direction can assist both cognitive and computational
investigations of moral judgments. | cs.CL | ClarifyDelphi: Reinforced Clarification Questions with Defeasibility Rewards for Social and Moral Situations |
2022-12-20T16:32:54Z | http://arxiv.org/pdf/2212.10408v1 | http://arxiv.org/abs/2212.10408v1 | Fahim Faisal, Antonios Anastasopoulos | 2022-12-20T16:32:54Z | cs.CL | Pretrained language models (PLMs) often fail to fairly represent target users
from certain world regions because of the under-representation of those regions
in training datasets. With recent PLMs trained on enormous data sources,
quantifying their potential biases is difficult, due to their black-box nature
and the sheer scale of the data sources. In this work, we devise an approach to
study the geographic bias (and knowledge) present in PLMs, proposing a
Geographic-Representation Probing Framework adopting a self-conditioning method
coupled with entity-country mappings. Our findings suggest PLMs'
representations map surprisingly well to the physical world in terms of
country-to-country associations, but this knowledge is unequally shared across
languages. Last, we explain how large PLMs despite exhibiting notions of
geographical proximity, over-amplify geopolitical favouritism at inference
time. | cs.CL | Geographic and Geopolitical Biases of Language Models |
2022-12-20T16:30:11Z | http://arxiv.org/pdf/2212.10405v2 | http://arxiv.org/abs/2212.10405v2 | Wenjie Yin, Vibhor Agarwal, Aiqi Jiang, Arkaitz Zubiaga, Nishanth Sastry | 2023-01-10T14:48:54Z | cs.CL, cs.SI | Supervised approaches generally rely on majority-based labels. However, it is
hard to achieve high agreement among annotators in subjective tasks such as
hate speech detection. Existing neural network models principally regard labels
as categorical variables, while ignoring the semantic information in diverse
label texts. In this paper, we propose AnnoBERT, a first-of-its-kind
architecture integrating annotator characteristics and label text with a
transformer-based model to detect hate speech, with unique representations
based on each annotator's characteristics via Collaborative Topic Regression
(CTR) and integrate label text to enrich textual representations. During
training, the model associates annotators with their label choices given a
piece of text; during evaluation, when label information is not available, the
model predicts the aggregated label given by the participating annotators by
utilising the learnt association. The proposed approach displayed an advantage
in detecting hate speech, especially in the minority class and edge cases with
annotator disagreement. Improvement in the overall performance is the largest
when the dataset is more label-imbalanced, suggesting its practical value in
identifying real-world hate speech, as the volume of hate speech in-the-wild is
extremely small on social media, when compared with normal (non-hate) speech.
Through ablation studies, we show the relative contributions of annotator
embeddings and label text to the model performance, and tested a range of
alternative annotator embeddings and label text combinations. | cs.CL | AnnoBERT: Effectively Representing Multiple Annotators' Label Choices to Improve Hate Speech Detection |
2022-12-20T16:29:03Z | http://arxiv.org/pdf/2212.10403v2 | http://arxiv.org/abs/2212.10403v2 | Jie Huang, Kevin Chen-Chuan Chang | 2023-05-26T17:59:33Z | cs.CL, cs.AI | Reasoning is a fundamental aspect of human intelligence that plays a crucial
role in activities such as problem solving, decision making, and critical
thinking. In recent years, large language models (LLMs) have made significant
progress in natural language processing, and there is observation that these
models may exhibit reasoning abilities when they are sufficiently large.
However, it is not yet clear to what extent LLMs are capable of reasoning. This
paper provides a comprehensive overview of the current state of knowledge on
reasoning in LLMs, including techniques for improving and eliciting reasoning
in these models, methods and benchmarks for evaluating reasoning abilities,
findings and implications of previous research in this field, and suggestions
on future directions. Our aim is to provide a detailed and up-to-date review of
this topic and stimulate meaningful discussion and future work. | cs.CL | Towards Reasoning in Large Language Models: A Survey |
2022-12-20T16:26:18Z | http://arxiv.org/pdf/2212.10400v1 | http://arxiv.org/abs/2212.10400v1 | Weiwei Sun, Zhengliang Shi, Shen Gao, Pengjie Ren, Maarten de Rijke, Zhaochun Ren | 2022-12-20T16:26:18Z | cs.CL, cs.AI | Pre-trained language models (LMs) store knowledge in their parameters and can
generate informative responses when used in conversational systems. However,
LMs suffer from the problem of "hallucination:" they may generate
plausible-looking statements that are irrelevant or factually incorrect. To
address this problem, we propose a contrastive learning scheme, named MixCL. A
novel mixed contrastive objective is proposed to explicitly optimize the
implicit knowledge elicitation process of LMs, and thus reduce their
hallucination in conversations. We also examine negative sampling strategies of
retrieved hard negatives and model-generated negatives. We conduct experiments
on Wizard-of-Wikipedia, a public, open-domain knowledge-grounded dialogue
benchmark, and assess the effectiveness of MixCL. MixCL effectively reduces the
hallucination of LMs in conversations and achieves the highest performance
among LM-based dialogue agents in terms of relevancy and factuality. We show
that MixCL achieves comparable performance to state-of-the-art KB-based
approaches while enjoying notable advantages in terms of efficiency and
scalability. | cs.CL | Contrastive Learning Reduces Hallucination in Conversations |
2022-12-20T16:25:42Z | http://arxiv.org/pdf/2212.10397v3 | http://arxiv.org/abs/2212.10397v3 | Lining Zhang, Simon Mille, Yufang Hou, Daniel Deutsch, Elizabeth Clark, Yixin Liu, Saad Mahamood, Sebastian Gehrmann, Miruna Clinciu, Khyathi Chandu, João Sedoc | 2023-06-14T01:45:33Z | cs.CL | To prevent the costly and inefficient use of resources on low-quality
annotations, we want a method for creating a pool of dependable annotators who
can effectively complete difficult tasks, such as evaluating automatic
summarization. Thus, we investigate the recruitment of high-quality Amazon
Mechanical Turk workers via a two-step pipeline. We show that we can
successfully filter out subpar workers before they carry out the evaluations
and obtain high-agreement annotations with similar constraints on resources.
Although our workers demonstrate a strong consensus among themselves and
CloudResearch workers, their alignment with expert judgments on a subset of the
data is not as expected and needs further training in correctness. This paper
still serves as a best practice for the recruitment of qualified annotators in
other challenging annotation tasks. | cs.CL | Needle in a Haystack: An Analysis of High-Agreement Workers on MTurk for Summarization |
2022-12-20T16:20:56Z | http://arxiv.org/pdf/2212.10392v1 | http://arxiv.org/abs/2212.10392v1 | Jianhua Yuan, Yanyan Zhao, Bing Qin | 2022-12-20T16:20:56Z | cs.CL | Stance detection models may tend to rely on dataset bias in the text part as
a shortcut and thus fail to sufficiently learn the interaction between the
targets and texts. Recent debiasing methods usually treated features learned by
small models or big models at earlier steps as bias features and proposed to
exclude the branch learning those bias features during inference. However, most
of these methods fail to disentangle the ``good'' stance features and ``bad''
bias features in the text part. In this paper, we investigate how to mitigate
dataset bias in stance detection. Motivated by causal effects, we leverage a
novel counterfactual inference framework, which enables us to capture the
dataset bias in the text part as the direct causal effect of the text on
stances and reduce the dataset bias in the text part by subtracting the direct
text effect from the total causal effect. We novelly model bias features as
features that correlate with the stance labels but fail on intermediate stance
reasoning subtasks and propose an adversarial bias learning module to model the
bias more accurately. To verify whether our model could better model the
interaction between texts and targets, we test our model on recently proposed
test sets to evaluate the understanding of the task from various aspects.
Experiments demonstrate that our proposed method (1) could better model the
bias features, and (2) outperforms existing debiasing baselines on both the
original dataset and most of the newly constructed test sets. | cs.CL | Debiasing Stance Detection Models with Counterfactual Reasoning and Adversarial Bias Learning |
2022-12-20T16:18:03Z | http://arxiv.org/pdf/2212.10391v2 | http://arxiv.org/abs/2212.10391v2 | Jimin Hong, Jungsoo Park, Daeyoung Kim, Seongjae Choi, Bokyung Son, Jaewook Kang | 2023-05-19T08:37:33Z | cs.CL | With contrastive pre-training, sentence encoders are generally optimized to
locate semantically similar samples closer to each other in their embedding
spaces. In this work, we focus on the potential of their embedding spaces to be
readily adapted to zero-shot text classification, as semantically distinct
samples are already well-separated. Our framework, RaLP (Retrieval augmented
Label Prompts for sentence encoder), encodes prompted label candidates with a
sentence encoder, then assigns the label whose prompt embedding has the highest
similarity with the input text embedding. In order to compensate for the
potentially poorly descriptive labels in their original format, RaLP retrieves
sentences that are semantically similar to the original label prompt from
external corpora and use them as additional pseudo-label prompts. RaLP achieves
competitive or stronger performance than much larger baselines on various
closed-set classification and multiple-choice QA datasets under zero-shot
settings. We show that the retrieval component plays a pivotal role in RaLP's
success, and its results are robustly attained regardless of verbalizer
variations. | cs.CL | Empowering Sentence Encoders with Prompting and Label Retrieval for Zero-shot Text Classification |
2022-12-20T16:06:09Z | http://arxiv.org/pdf/2212.10381v1 | http://arxiv.org/abs/2212.10381v1 | Dheeru Dua, Emma Strubell, Sameer Singh, Pat Verga | 2022-12-20T16:06:09Z | cs.CL | Recent advances in open-domain question answering (ODQA) have demonstrated
impressive accuracy on standard Wikipedia style benchmarks. However, it is less
clear how robust these models are and how well they perform when applied to
real-world applications in drastically different domains. While there has been
some work investigating how well ODQA models perform when tested for
out-of-domain (OOD) generalization, these studies have been conducted only
under conservative shifts in data distribution and typically focus on a single
component (ie. retrieval) rather than an end-to-end system. In response, we
propose a more realistic and challenging domain shift evaluation setting and,
through extensive experiments, study end-to-end model performance. We find that
not only do models fail to generalize, but high retrieval scores often still
yield poor answer prediction accuracy. We then categorize different types of
shifts and propose techniques that, when presented with a new dataset, predict
if intervention methods are likely to be successful. Finally, using insights
from this analysis, we propose and evaluate several intervention methods which
improve end-to-end answer F1 score by up to 24 points. | cs.CL | To Adapt or to Annotate: Challenges and Interventions for Domain Adaptation in Open-Domain Question Answering |
2022-12-20T16:03:25Z | http://arxiv.org/pdf/2212.10380v2 | http://arxiv.org/abs/2212.10380v2 | Ori Ram, Liat Bezalel, Adi Zicher, Yonatan Belinkov, Jonathan Berant, Amir Globerson | 2023-05-24T12:05:27Z | cs.CL, cs.IR | Dual encoders are now the dominant architecture for dense retrieval. Yet, we
have little understanding of how they represent text, and why this leads to
good performance. In this work, we shed light on this question via
distributions over the vocabulary. We propose to interpret the vector
representations produced by dual encoders by projecting them into the model's
vocabulary space. We show that the resulting projections contain rich semantic
information, and draw connection between them and sparse retrieval. We find
that this view can offer an explanation for some of the failure cases of dense
retrievers. For example, we observe that the inability of models to handle tail
entities is correlated with a tendency of the token distributions to forget
some of the tokens of those entities. We leverage this insight and propose a
simple way to enrich query and passage representations with lexical information
at inference time, and show that this significantly improves performance
compared to the original model in zero-shot settings, and specifically on the
BEIR benchmark. | cs.CL | What Are You Token About? Dense Retrieval as Distributions Over the Vocabulary |
2022-12-20T15:58:54Z | http://arxiv.org/pdf/2212.10378v2 | http://arxiv.org/abs/2212.10378v2 | Ting-Yun Chang, Robin Jia | 2023-05-24T22:32:56Z | cs.CL | In-context learning (ICL) enables large language models (LLMs) to perform new
tasks by prompting them with a sequence of training examples. However, it is
known that ICL is very sensitive to the choice of training examples: randomly
sampling examples from a training set leads to high variance in performance. In
this paper, we show that carefully curating a subset of training data greatly
stabilizes ICL performance without any other changes to the ICL algorithm
(e.g., prompt retrieval or calibration). We introduce two methods to choose
training subsets -- both score training examples individually, then select the
highest-scoring ones. CondAcc scores a training example by its average dev-set
ICL accuracy when combined with random training examples, while Datamodels
learns linear regressors that estimate how the presence of each training
example influences LLM outputs. Across five tasks and two LLMs, sampling from
stable subsets selected by CondAcc and Datamodels improves average accuracy
over sampling from the entire training set by 7.7% and 6.3%, respectively.
Surprisingly, the stable subset examples are not especially diverse in content
or low in perplexity, in contrast with other work suggesting that diversity and
perplexity are important when prompting LLMs. | cs.CL | Data Curation Alone Can Stabilize In-context Learning |
2022-12-20T15:55:21Z | http://arxiv.org/pdf/2212.10375v2 | http://arxiv.org/abs/2212.10375v2 | Zhiyong Wu, Yaoxiang Wang, Jiacheng Ye, Lingpeng Kong | 2023-05-03T14:43:50Z | cs.CL, cs.AI | Despite the surprising few-shot performance of in-context learning (ICL), it
is still a common practice to randomly sample examples to serve as context.
This paper advocates a new principle for ICL: self-adaptive in-context
learning. The self-adaption mechanism is introduced to help each sample find an
in-context example permutation (i.e., selection and ordering) that can derive
the correct prediction, thus maximizing performance. To validate the
effectiveness of self-adaptive ICL, we propose a general select-then-rank
framework and instantiate it with new selection and ranking algorithms. Upon
extensive evaluation on eight different NLP datasets, our self-adaptive ICL
method achieves a 40% relative improvement over the common practice setting.
Further analysis reveals the enormous potential of self-adaptive ICL that it
might be able to close the gap between ICL and finetuning given more advanced
algorithms. Our code is released to facilitate future research in this area:
https://github.com/Shark-NLP/self-adaptive-ICL | cs.CL | Self-Adaptive In-Context Learning: An Information Compression Perspective for In-Context Example Selection and Ordering |
2022-12-20T15:40:17Z | http://arxiv.org/pdf/2212.10356v2 | http://arxiv.org/abs/2212.10356v2 | Ta-Chung Chi, Ting-Han Fan, Alexander I. Rudnicky, Peter J. Ramadge | 2023-05-23T21:18:09Z | cs.CL | Length extrapolation permits training a transformer language model on short
sequences that preserves perplexities when tested on substantially longer
sequences. A relative positional embedding design, ALiBi, has had the widest
usage to date. We dissect ALiBi via the lens of receptive field analysis
empowered by a novel cumulative normalized gradient tool. The concept of
receptive field further allows us to modify the vanilla Sinusoidal positional
embedding to create ~\textbf{Sandwich}, the first parameter-free relative
positional embedding design that truly length information uses longer than the
training sequence. Sandwich shares with KERPLE and T5 the same logarithmic
decaying temporal bias pattern with learnable relative positional embeddings;
these elucidate future extrapolatable positional embedding design. | cs.CL | Dissecting Transformer Length Extrapolation via the Lens of Receptive Field Analysis |
2022-12-20T15:31:58Z | http://arxiv.org/pdf/2212.10346v1 | http://arxiv.org/abs/2212.10346v1 | Matthias Kraus, Diana Betancourt, Wolfgang Minker | 2022-12-20T15:31:58Z | cs.RO, cs.CL | Using robots in educational contexts has already shown to be beneficial for a
student's learning and social behaviour. For levitating them to the next level
of providing more effective and human-like tutoring, the ability to adapt to
the user and to express proactivity is fundamental. By acting proactively,
intelligent robotic tutors anticipate possible situations where problems for
the student may arise and act in advance for preventing negative outcomes.
Still, the decisions of when and how to behave proactively are open questions.
Therefore, this paper deals with the investigation of how the student's
cognitive-affective states can be used by a robotic tutor for triggering
proactive tutoring dialogue. In doing so, it is aimed to improve the learning
experience. For this reason, a concept learning task scenario was observed
where a robotic assistant proactively helped when negative user states were
detected. In a learning task, the user's states of frustration and confusion
were deemed to have negative effects on the outcome of the task and were used
to trigger proactive behaviour. In an empirical user study with 40
undergraduate and doctoral students, we studied whether the initiation of
proactive behaviour after the detection of signs of confusion and frustration
improves the student's concentration and trust in the agent. Additionally, we
investigated which level of proactive dialogue is useful for promoting the
student's concentration and trust. The results show that high proactive
behaviour harms trust, especially when triggered during negative
cognitive-affective states but contributes to keeping the student focused on
the task when triggered in these states. Based on our study results, we further
discuss future steps for improving the proactive assistance of robotic tutoring
systems. | cs.RO | Does It Affect You? Social and Learning Implications of Using Cognitive-Affective State Recognition for Proactive Human-Robot Tutoring |
2022-12-20T15:26:19Z | http://arxiv.org/pdf/2212.10341v2 | http://arxiv.org/abs/2212.10341v2 | Xiaoming Liu, Zhaohan Zhang, Yichen Wang, Hang Pu, Yu Lan, Chao Shen | 2023-10-20T13:21:51Z | cs.CL | Machine-Generated Text (MGT) detection, a task that discriminates MGT from
Human-Written Text (HWT), plays a crucial role in preventing misuse of text
generative models, which excel in mimicking human writing style recently.
Latest proposed detectors usually take coarse text sequences as input and
fine-tune pretrained models with standard cross-entropy loss. However, these
methods fail to consider the linguistic structure of texts. Moreover, they lack
the ability to handle the low-resource problem which could often happen in
practice considering the enormous amount of textual data online. In this paper,
we present a coherence-based contrastive learning model named CoCo to detect
the possible MGT under low-resource scenario. To exploit the linguistic
feature, we encode coherence information in form of graph into text
representation. To tackle the challenges of low data resource, we employ a
contrastive learning framework and propose an improved contrastive loss for
preventing performance degradation brought by simple samples. The experiment
results on two public datasets and two self-constructed datasets prove our
approach outperforms the state-of-art methods significantly. Also, we
surprisingly find that MGTs originated from up-to-date language models could be
easier to detect than these from previous models, in our experiments. And we
propose some preliminary explanations for this counter-intuitive phenomena. All
the codes and datasets are open-sourced. | cs.CL | CoCo: Coherence-Enhanced Machine-Generated Text Detection Under Data Limitation With Contrastive Learning |
2022-12-20T15:16:24Z | http://arxiv.org/pdf/2212.10325v5 | http://arxiv.org/abs/2212.10325v5 | Hongyi Yuan, Zheng Yuan, Chuanqi Tan, Fei Huang, Songfang Huang | 2023-05-22T17:31:46Z | cs.CL | Diffusion model, a new generative modelling paradigm, has achieved great
success in image, audio, and video generation. However, considering the
discrete categorical nature of text, it is not trivial to extend continuous
diffusion models to natural language, and text diffusion models are less
studied. Sequence-to-sequence text generation is one of the essential natural
language processing topics. In this work, we apply diffusion models to approach
sequence-to-sequence text generation, and explore whether the superiority
generation performance of diffusion model can transfer to natural language
domain. We propose SeqDiffuSeq, a text diffusion model for sequence-to-sequence
generation. SeqDiffuSeq uses an encoder-decoder Transformers architecture to
model denoising function. In order to improve generation quality, SeqDiffuSeq
combines the self-conditioning technique and a newly proposed adaptive noise
schedule technique. The adaptive noise schedule has the difficulty of denoising
evenly distributed across time steps, and considers exclusive noise schedules
for tokens at different positional order. Experiment results illustrate the
good performance on sequence-to-sequence generation in terms of text quality
and inference time. | cs.CL | SeqDiffuSeq: Text Diffusion with Encoder-Decoder Transformers |
2022-12-20T15:07:37Z | http://arxiv.org/pdf/2212.10315v2 | http://arxiv.org/abs/2212.10315v2 | Hamish Ivison, Akshita Bhagia, Yizhong Wang, Hannaneh Hajishirzi, Matthew Peters | 2023-05-24T22:30:30Z | cs.CL | Recent NLP models have shown the remarkable ability to effectively generalise
`zero-shot' to new tasks using only natural language instructions as guidance.
However, many of these approaches suffer from high computational costs due to
their reliance on concatenating lengthy instructions with every input example,
resulting in costly reprocessing of the instruction. To avoid this, we
introduce Hypernetworks for INstruction Tuning (HINT), which convert task
instructions and examples into parameter-efficient modules inserted into an
underlying model using a pretrained text encoder, eliminating the need to
include instructions in the model input. The hypernetwork in HINT also produces
an encoded instruction, which we concatenate with encoded inputs during
decoding to further improve performance. HINT models outperform strong
state-of-the-art baselines by over 10% when controlling for compute (measured
in FLOPs). By converting instructions into modules, HINT models can effectively
disregard the length of instructions and few-shot example inputs in terms of
compute usage. As a result, HINT can enhance its performance by up to 25% by
incorporating additional few-shot data, while utilizing only up to 5% more
compute. This combines the strengths of parameter-efficient fine-tuning and
in-context learning. | cs.CL | HINT: Hypernetwork Instruction Tuning for Efficient Zero- & Few-Shot Generalisation |
2022-12-20T15:02:38Z | http://arxiv.org/pdf/2212.10313v2 | http://arxiv.org/abs/2212.10313v2 | Yaoming Zhu, Zewei Sun, Shanbo Cheng, Luyang Huang, Liwei Wu, Mingxuan Wang | 2023-09-03T03:46:05Z | cs.CL | Multimodal machine translation (MMT) aims to improve translation quality by
incorporating information from other modalities, such as vision. Previous MMT
systems mainly focus on better access and use of visual information and tend to
validate their methods on image-related datasets. These studies face two
challenges. First, they can only utilize triple data (bilingual texts with
images), which is scarce; second, current benchmarks are relatively restricted
and do not correspond to realistic scenarios. Therefore, this paper
correspondingly establishes new methods and new datasets for MMT. First, we
propose a framework 2/3-Triplet with two new approaches to enhance MMT by
utilizing large-scale non-triple data: monolingual image-text data and parallel
text-only data. Second, we construct an English-Chinese {e}-commercial
{m}ulti{m}odal {t}ranslation dataset (including training and testing), named
EMMT, where its test set is carefully selected as some words are ambiguous and
shall be translated mistakenly without the help of images. Experiments show
that our method is more suitable for real-world scenarios and can significantly
improve translation performance by using more non-triple data. In addition, our
model also rivals various SOTA models in conventional multimodal translation
benchmarks. | cs.CL | Beyond Triplet: Leveraging the Most Data for Multimodal Machine Translation |
2022-12-20T14:39:58Z | http://arxiv.org/pdf/2212.10297v2 | http://arxiv.org/abs/2212.10297v2 | Nikita Moghe, Tom Sherborne, Mark Steedman, Alexandra Birch | 2023-06-19T03:59:55Z | cs.CL, cs.AI | Automatic machine translation (MT) metrics are widely used to distinguish the
translation qualities of machine translation systems across relatively large
test sets (system-level evaluation). However, it is unclear if automatic
metrics are reliable at distinguishing good translations from bad translations
at the sentence level (segment-level evaluation). In this paper, we investigate
how useful MT metrics are at detecting the success of a machine translation
component when placed in a larger platform with a downstream task. We evaluate
the segment-level performance of the most widely used MT metrics (chrF, COMET,
BERTScore, etc.) on three downstream cross-lingual tasks (dialogue state
tracking, question answering, and semantic parsing). For each task, we only
have access to a monolingual task-specific model. We calculate the correlation
between the metric's ability to predict a good/bad translation with the
success/failure on the final task for the Translate-Test setup. Our experiments
demonstrate that all metrics exhibit negligible correlation with the extrinsic
evaluation of the downstream outcomes. We also find that the scores provided by
neural metrics are not interpretable mostly because of undefined ranges. We
synthesise our analysis into recommendations for future MT metrics to produce
labels rather than scores for more informative interaction between machine
translation and multilingual language understanding. | cs.CL | Extrinsic Evaluation of Machine Translation Metrics |
2022-12-20T14:11:31Z | http://arxiv.org/pdf/2212.10264v1 | http://arxiv.org/abs/2212.10264v1 | Shiqi Wang, Zheng Li, Haifeng Qian, Chenghao Yang, Zijian Wang, Mingyue Shang, Varun Kumar, Samson Tan, Baishakhi Ray, Parminder Bhatia, Ramesh Nallapati, Murali Krishna Ramanathan, Dan Roth, Bing Xiang | 2022-12-20T14:11:31Z | cs.LG, cs.CL, cs.SE | Code generation models have achieved impressive performance. However, they
tend to be brittle as slight edits to a prompt could lead to very different
generations; these robustness properties, critical for user experience when
deployed in real-life applications, are not well understood. Most existing
works on robustness in text or code tasks have focused on classification, while
robustness in generation tasks is an uncharted area and to date there is no
comprehensive benchmark for robustness in code generation. In this paper, we
propose ReCode, a comprehensive robustness evaluation benchmark for code
generation models. We customize over 30 transformations specifically for code
on docstrings, function and variable names, code syntax, and code format. They
are carefully designed to be natural in real-life coding practice, preserve the
original semantic meaning, and thus provide multifaceted assessments of a
model's robustness performance. With human annotators, we verified that over
90% of the perturbed prompts do not alter the semantic meaning of the original
prompt. In addition, we define robustness metrics for code generation models
considering the worst-case behavior under each type of perturbation, taking
advantage of the fact that executing the generated code can serve as objective
evaluation. We demonstrate ReCode on SOTA models using HumanEval, MBPP, as well
as function completion tasks derived from them. Interesting observations
include: better robustness for CodeGen over InCoder and GPT-J; models are most
sensitive to syntax perturbations; more challenging robustness evaluation on
MBPP over HumanEval. | cs.LG | ReCode: Robustness Evaluation of Code Generation Models |
2022-12-20T14:06:50Z | http://arxiv.org/pdf/2212.10258v2 | http://arxiv.org/abs/2212.10258v2 | Yahan Yang, Soham Dan, Dan Roth, Insup Lee | 2023-07-11T19:33:44Z | cs.CL, cs.CR, cs.LG | Recently it has been shown that state-of-the-art NLP models are vulnerable to
adversarial attacks, where the predictions of a model can be drastically
altered by slight modifications to the input (such as synonym substitutions).
While several defense techniques have been proposed, and adapted, to the
discrete nature of text adversarial attacks, the benefits of general-purpose
regularization methods such as label smoothing for language models, have not
been studied. In this paper, we study the adversarial robustness provided by
various label smoothing strategies in foundational models for diverse NLP tasks
in both in-domain and out-of-domain settings. Our experiments show that label
smoothing significantly improves adversarial robustness in pre-trained models
like BERT, against various popular attacks. We also analyze the relationship
between prediction confidence and robustness, showing that label smoothing
reduces over-confident errors on adversarial examples. | cs.CL | In and Out-of-Domain Text Adversarial Robustness via Label Smoothing |
2022-12-20T14:06:45Z | http://arxiv.org/pdf/2212.10257v1 | http://arxiv.org/abs/2212.10257v1 | Baopu Qiu, Liang Ding, Di Wu, Lin Shang, Yibing Zhan, Dacheng Tao | 2022-12-20T14:06:45Z | cs.CL | Machine Translation Quality Estimation (QE) is the task of evaluating
translation output in the absence of human-written references. Due to the
scarcity of human-labeled QE data, previous works attempted to utilize the
abundant unlabeled parallel corpora to produce additional training data with
pseudo labels. In this paper, we demonstrate a significant gap between parallel
data and real QE data: for QE data, it is strictly guaranteed that the source
side is original texts and the target side is translated (namely
translationese). However, for parallel data, it is indiscriminate and the
translationese may occur on either source or target side. We compare the impact
of parallel data with different translation directions in QE data augmentation,
and find that using the source-original part of parallel corpus consistently
outperforms its target-original counterpart. Moreover, since the WMT corpus
lacks direction information for each parallel sentence, we train a classifier
to distinguish source- and target-original bitext, and carry out an analysis of
their difference in both style and domain. Together, these findings suggest
using source-original parallel data for QE data augmentation, which brings a
relative improvement of up to 4.0% and 6.4% compared to undifferentiated data
on sentence- and word-level QE tasks respectively. | cs.CL | Original or Translated? On the Use of Parallel Data for Translation Quality Estimation |
2022-12-20T13:36:25Z | http://arxiv.org/pdf/2212.10240v2 | http://arxiv.org/abs/2212.10240v2 | Lihua Qian, Mingxuan Wang, Yang Liu, Hao Zhou | 2023-11-29T08:30:34Z | cs.CL | Previously, non-autoregressive models were widely perceived as being superior
in generation efficiency but inferior in generation quality due to the
difficulties of modeling multiple target modalities. To enhance the
multi-modality modeling ability, we propose the diffusion glancing transformer,
which employs a modality diffusion process and residual glancing sampling. The
modality diffusion process is a discrete process that interpolates the
multi-modal distribution along the decoding steps, and the residual glancing
sampling approach guides the model to continuously learn the remaining
modalities across the layers. Experimental results on various machine
translation and text generation benchmarks demonstrate that DIFFGLAT achieves
better generation accuracy while maintaining fast decoding speed compared with
both autoregressive and non-autoregressive models. | cs.CL | Diffusion Glancing Transformer for Parallel Sequence to Sequence Learning |
2022-12-20T13:20:21Z | http://arxiv.org/pdf/2212.10233v2 | http://arxiv.org/abs/2212.10233v2 | Di Wu, Wasi Uddin Ahmad, Kai-Wei Chang | 2024-02-23T04:45:40Z | cs.CL | Neural models that do not rely on pre-training have excelled in the keyphrase
generation task with large annotated datasets. Meanwhile, new approaches have
incorporated pre-trained language models (PLMs) for their data efficiency.
However, there lacks a systematic study of how the two types of approaches
compare and how different design choices can affect the performance of
PLM-based models. To fill in this knowledge gap and facilitate a more informed
use of PLMs for keyphrase extraction and keyphrase generation, we present an
in-depth empirical study. Formulating keyphrase extraction as sequence labeling
and keyphrase generation as sequence-to-sequence generation, we perform
extensive experiments in three domains. After showing that PLMs have
competitive high-resource performance and state-of-the-art low-resource
performance, we investigate important design choices including in-domain PLMs,
PLMs with different pre-training objectives, using PLMs with a parameter
budget, and different formulations for present keyphrases. Further results show
that (1) in-domain BERT-like PLMs can be used to build strong and
data-efficient keyphrase generation models; (2) with a fixed parameter budget,
prioritizing model depth over width and allocating more layers in the encoder
leads to better encoder-decoder models; and (3) introducing four in-domain
PLMs, we achieve a competitive performance in the news domain and the
state-of-the-art performance in the scientific domain. | cs.CL | Pre-trained Language Models for Keyphrase Generation: A Thorough Empirical Study |
2022-12-20T12:51:11Z | http://arxiv.org/pdf/2212.10218v2 | http://arxiv.org/abs/2212.10218v2 | Jian Yang, Shuming Ma, Li Dong, Shaohan Huang, Haoyang Huang, Yuwei Yin, Dongdong Zhang, Liqun Yang, Furu Wei, Zhoujun Li | 2023-05-09T09:08:33Z | cs.CL | Pre-trained models have achieved remarkable success in natural language
processing (NLP). However, existing pre-training methods underutilize the
benefits of language understanding for generation. Inspired by the idea of
Generative Adversarial Networks (GANs), we propose a GAN-style model for
encoder-decoder pre-training by introducing an auxiliary discriminator,
unifying the ability of language understanding and generation in a single
model. Our model, named as GanLM, is trained with two pre-training objectives:
replaced token detection and replaced token denoising. Specifically, given
masked source sentences, the generator outputs the target distribution and the
discriminator predicts whether the target sampled tokens from distribution are
incorrect. The target sentence is replaced with misclassified tokens to
construct noisy previous context, which is used to generate the gold sentence.
In general, both tasks improve the ability of language understanding and
generation by selectively using the denoising data. Extensive experiments in
language generation benchmarks show that GanLM with the powerful language
understanding capability outperforms various strong pre-trained language models
(PLMs) and achieves state-of-the-art performance. | cs.CL | GanLM: Encoder-Decoder Pre-training with an Auxiliary Discriminator |
2022-12-20T12:16:46Z | http://arxiv.org/pdf/2212.10197v2 | http://arxiv.org/abs/2212.10197v2 | Tong Zheng, Bei Li, Huiwen Bao, Tong Xiao, Jingbo Zhu | 2024-06-05T10:18:14Z | cs.CL | Two principles: the complementary principle and the consensus principle are
widely acknowledged in the literature of multi-view learning. However, the
current design of multi-head self-attention, an instance of multi-view
learning, prioritizes the complementarity while ignoring the consensus. To
address this problem, we propose an enhanced multi-head self-attention (EMHA).
First, to satisfy the complementary principle, EMHA removes the one-to-one
mapping constraint among queries and keys in multiple subspaces and allows each
query to attend to multiple keys. On top of that, we develop a method to fully
encourage consensus among heads by introducing two interaction models, namely
inner-subspace interaction and cross-subspace interaction. Extensive
experiments on a wide range of language tasks (e.g., machine translation,
abstractive summarization and grammar correction, language modeling), show its
superiority, with a very modest increase in model size. Our code would be
available at: https://github.com/zhengkid/EIT-Enhanced-Interactive-Transformer. | cs.CL | EIT: Enhanced Interactive Transformer |
2022-12-20T12:03:19Z | http://arxiv.org/pdf/2212.10192v2 | http://arxiv.org/abs/2212.10192v2 | Chongyang Tao, Chang Liu, Tao Shen, Can Xu, Xiubo Geng, Binxing Jiao, Daxin Jiang | 2024-06-06T15:20:27Z | cs.CL | To improve the performance of the dual-encoder retriever, one effective
approach is knowledge distillation from the cross-encoder ranker. Existing
works construct the candidate passages following the supervised learning
setting where a query is paired with a positive passage and a batch of
negatives. However, through empirical observation, we find that even the hard
negatives from advanced methods are still too trivial for the teacher to
distinguish, preventing the teacher from transferring abundant dark knowledge
to the student through its soft label. To alleviate this issue, we propose
ADAM, a knowledge distillation framework that can better transfer the dark
knowledge held in the teacher with Adaptive Dark exAMples. Different from
previous works that only rely on one positive and hard negatives as candidate
passages, we create dark examples that all have moderate relevance to the query
through mixing-up and masking in discrete space. Furthermore, as the quality of
knowledge held in different training instances varies as measured by the
teacher's confidence score, we propose a self-paced distillation strategy that
adaptively concentrates on a subset of high-quality instances to conduct our
dark-example-based knowledge distillation to help the student learn better. We
conduct experiments on two widely-used benchmarks and verify the effectiveness
of our method. | cs.CL | Adam: Dense Retrieval Distillation with Adaptive Dark Examples |
2022-12-20T12:02:40Z | http://arxiv.org/pdf/2212.10191v1 | http://arxiv.org/abs/2212.10191v1 | Tao Wang, Jiangyan Yi, Ruibo Fu, Jianhua Tao, Zhengqi Wen, Chu Yuan Zhang | 2022-12-20T12:02:40Z | cs.SD, cs.CL, cs.LG, eess.AS | Text-based speech editing allows users to edit speech by intuitively cutting,
copying, and pasting text to speed up the process of editing speech. In the
previous work, CampNet (context-aware mask prediction network) is proposed to
realize text-based speech editing, significantly improving the quality of
edited speech. This paper aims at a new task: adding emotional effect to the
editing speech during the text-based speech editing to make the generated
speech more expressive. To achieve this task, we propose Emo-CampNet (emotion
CampNet), which can provide the option of emotional attributes for the
generated speech in text-based speech editing and has the one-shot ability to
edit unseen speakers' speech. Firstly, we propose an end-to-end
emotion-selectable text-based speech editing model. The key idea of the model
is to control the emotion of generated speech by introducing additional emotion
attributes based on the context-aware mask prediction network. Secondly, to
prevent the emotion of the generated speech from being interfered by the
emotional components in the original speech, a neutral content generator is
proposed to remove the emotion from the original speech, which is optimized by
the generative adversarial framework. Thirdly, two data augmentation methods
are proposed to enrich the emotional and pronunciation information in the
training set, which can enable the model to edit the unseen speaker's speech.
The experimental results that 1) Emo-CampNet can effectively control the
emotion of the generated speech in the process of text-based speech editing;
And can edit unseen speakers' speech. 2) Detailed ablation experiments further
prove the effectiveness of emotional selectivity and data augmentation methods.
The demo page is available at https://hairuo55.github.io/Emo-CampNet/ | cs.SD | Emotion Selectable End-to-End Text-based Speech Editing |
2022-12-20T12:02:34Z | http://arxiv.org/pdf/2212.10190v1 | http://arxiv.org/abs/2212.10190v1 | Xun Wang, Tao Ge, Allen Mao, Yuki Li, Furu Wei, Si-Qing Chen | 2022-12-20T12:02:34Z | cs.CL | We introduce \textsc{PoliteRewrite} -- a dataset for polite language rewrite
which is a novel sentence rewrite task. Compared with previous text style
transfer tasks that can be mostly addressed by slight token- or phrase-level
edits, polite language rewrite requires deep understanding and extensive
sentence-level edits over an offensive and impolite sentence to deliver the
same message euphemistically and politely, which is more challenging -- not
only for NLP models but also for human annotators to rewrite with effort. To
alleviate the human effort for efficient annotation, we first propose a novel
annotation paradigm by a collaboration of human annotators and GPT-3.5 to
annotate \textsc{PoliteRewrite}. The released dataset has 10K polite sentence
rewrites annotated collaboratively by GPT-3.5 and human, which can be used as
gold standard for training, validation and test; and 100K high-quality polite
sentence rewrites by GPT-3.5 without human review. We wish this work (The
dataset (10K+100K) will be released soon) could contribute to the research on
more challenging sentence rewrite, and provoke more thought in future on
resource annotation paradigm with the help of the large-scaled pretrained
models. | cs.CL | Pay Attention to Your Tone: Introducing a New Dataset for Polite Language Rewrite |
2022-12-20T12:00:26Z | http://arxiv.org/pdf/2212.10189v2 | http://arxiv.org/abs/2212.10189v2 | Mayur Patidar, Prayushi Faldu, Avinash Singh, Lovekesh Vig, Indrajit Bhattacharya, Mausam | 2023-06-24T11:06:55Z | cs.CL, cs.AI | When answering natural language questions over knowledge bases, missing
facts, incomplete schema and limited scope naturally lead to many questions
being unanswerable. While answerability has been explored in other QA settings,
it has not been studied for QA over knowledge bases (KBQA). We create
GrailQAbility, a new benchmark KBQA dataset with unanswerability, by first
identifying various forms of KB incompleteness that make questions
unanswerable, and then systematically adapting GrailQA (a popular KBQA dataset
with only answerable questions). Experimenting with three state-of-the-art KBQA
models, we find that all three models suffer a drop in performance even after
suitable adaptation for unanswerable questions. In addition, these often detect
unanswerability for wrong reasons and find specific forms of unanswerability
particularly difficult to handle. This underscores the need for further
research in making KBQA systems robust to unanswerability | cs.CL | Do I have the Knowledge to Answer? Investigating Answerability of Knowledge Base Questions |
2022-12-20T11:52:52Z | http://arxiv.org/pdf/2301.05150v1 | http://arxiv.org/abs/2301.05150v1 | Maksimjeet Chowdhary, Sanyam Goyal, Venktesh V, Mukesh Mohania, Vikram Goyal | 2022-12-20T11:52:52Z | cs.CL | Online learning platforms provide diverse questions to gauge the learners'
understanding of different concepts. The repository of questions has to be
constantly updated to ensure a diverse pool of questions to conduct assessments
for learners. However, it is impossible for the academician to manually skim
through the large repository of questions to check for duplicates when
onboarding new questions from external sources. Hence, we propose a tool QDup
in this paper that can surface near-duplicate and semantically related
questions without any supervised data. The proposed tool follows an
unsupervised hybrid pipeline of statistical and neural approaches for
incorporating different nuances in similarity for the task of question
duplicate detection. We demonstrate that QDup can detect near-duplicate
questions and also suggest related questions for practice with remarkable
accuracy and speed from a large repository of questions. The demo video of the
tool can be found at https://www.youtube.com/watch?v=loh0_-7XLW4. | cs.CL | Unsupervised Question Duplicate and Related Questions Detection in e-learning platforms |
2022-12-20T11:37:22Z | http://arxiv.org/pdf/2212.10180v2 | http://arxiv.org/abs/2212.10180v2 | Ananya B. Sai, Vignesh Nagarajan, Tanay Dixit, Raj Dabre, Anoop Kunchukuttan, Pratyush Kumar, Mitesh M. Khapra | 2023-07-03T14:26:38Z | cs.CL | The rapid growth of machine translation (MT) systems has necessitated
comprehensive studies to meta-evaluate evaluation metrics being used, which
enables a better selection of metrics that best reflect MT quality.
Unfortunately, most of the research focuses on high-resource languages, mainly
English, the observations for which may not always apply to other languages.
Indian languages, having over a billion speakers, are linguistically different
from English, and to date, there has not been a systematic study of evaluating
MT systems from English into Indian languages. In this paper, we fill this gap
by creating an MQM dataset consisting of 7000 fine-grained annotations,
spanning 5 Indian languages and 7 MT systems, and use it to establish
correlations between annotator scores and scores obtained using existing
automatic metrics. Our results show that pre-trained metrics, such as COMET,
have the highest correlations with annotator scores. Additionally, we find that
the metrics do not adequately capture fluency-based errors in Indian languages,
and there is a need to develop metrics focused on Indian languages. We hope
that our dataset and analysis will help promote further research in this area. | cs.CL | IndicMT Eval: A Dataset to Meta-Evaluate Machine Translation metrics for Indian Languages |
2022-12-20T11:36:22Z | http://arxiv.org/pdf/2212.10179v1 | http://arxiv.org/abs/2212.10179v1 | Qingyu Lu, Liang Ding, Liping Xie, Kanjian Zhang, Derek F. Wong, Dacheng Tao | 2022-12-20T11:36:22Z | cs.CL | The state-of-the-art language model-based automatic metrics, e.g. BARTScore,
benefiting from large-scale contextualized pre-training, have been successfully
used in a wide range of natural language generation (NLG) tasks, including
machine translation, text summarization, and data-to-text. Recent studies show
that considering both major errors (e.g. mistranslated tokens) and minor errors
(e.g. imperfections in fluency) can produce high-quality human judgments. This
inspires us to approach the final goal of the evaluation metrics (human-like
evaluations) by automatic error analysis. To this end, we augment BARTScore by
incorporating the human-like error analysis strategies, namely BARTScore++,
where the final score consists of both the evaluations of major errors and
minor errors. Experimental results show that BARTScore++ can consistently
improve the performance of vanilla BARTScore and outperform existing
top-scoring metrics in 20 out of 25 test settings. We hope our technique can
also be extended to other pre-trained model-based metrics. We will release our
code and scripts to facilitate the community. | cs.CL | Toward Human-Like Evaluation for Natural Language Generation with Error Analysis |
2022-12-20T11:23:04Z | http://arxiv.org/pdf/2212.10173v1 | http://arxiv.org/abs/2212.10173v1 | Machel Reid, Mikel Artetxe | 2022-12-20T11:23:04Z | cs.CL | While prior work has established that the use of parallel data is conducive
for cross-lingual learning, it is unclear if the improvements come from the
data itself, or if it is the modeling of parallel interactions that matters.
Exploring this, we examine the usage of unsupervised machine translation to
generate synthetic parallel data, and compare it to supervised machine
translation and gold parallel data. We find that even model generated parallel
data can be useful for downstream tasks, in both a general setting (continued
pretraining) as well as the task-specific setting (translate-train), although
our best results are still obtained using real parallel data. Our findings
suggest that existing multilingual models do not exploit the full potential of
monolingual data, and prompt the community to reconsider the traditional
categorization of cross-lingual learning approaches. | cs.CL | On the Role of Parallel Data in Cross-lingual Transfer Learning |
2022-12-20T11:17:52Z | http://arxiv.org/pdf/2212.10171v1 | http://arxiv.org/abs/2212.10171v1 | Ridong Han, Tao Peng, Benyou Wang, Lu Liu, Xiang Wan | 2022-12-20T11:17:52Z | cs.CL | Document-level relation extraction faces two overlooked challenges: long-tail
problem and multi-label problem. Previous work focuses mainly on obtaining
better contextual representations for entity pairs, hardly address the above
challenges. In this paper, we analyze the co-occurrence correlation of
relations, and introduce it into DocRE task for the first time. We argue that
the correlations can not only transfer knowledge between data-rich relations
and data-scarce ones to assist in the training of tailed relations, but also
reflect semantic distance guiding the classifier to identify semantically close
relations for multi-label entity pairs. Specifically, we use relation embedding
as a medium, and propose two co-occurrence prediction sub-tasks from both
coarse- and fine-grained perspectives to capture relation correlations.
Finally, the learned correlation-aware embeddings are used to guide the
extraction of relational facts. Substantial experiments on two popular DocRE
datasets are conducted, and our method achieves superior results compared to
baselines. Insightful analysis also demonstrates the potential of relation
correlations to address the above challenges. | cs.CL | Document-level Relation Extraction with Relation Correlations |
2022-12-20T11:15:24Z | http://arxiv.org/pdf/2212.10168v2 | http://arxiv.org/abs/2212.10168v2 | Arnav Mhaske, Harshit Kedia, Sumanth Doddapaneni, Mitesh M. Khapra, Pratyush Kumar, Rudra Murthy V, Anoop Kunchukuttan | 2023-05-28T06:26:45Z | cs.CL | We present, Naamapadam, the largest publicly available Named Entity
Recognition (NER) dataset for the 11 major Indian languages from two language
families. The dataset contains more than 400k sentences annotated with a total
of at least 100k entities from three standard entity categories (Person,
Location, and, Organization) for 9 out of the 11 languages. The training
dataset has been automatically created from the Samanantar parallel corpus by
projecting automatically tagged entities from an English sentence to the
corresponding Indian language translation. We also create manually annotated
testsets for 9 languages. We demonstrate the utility of the obtained dataset on
the Naamapadam-test dataset. We also release IndicNER, a multilingual IndicBERT
model fine-tuned on Naamapadam training set. IndicNER achieves an F1 score of
more than $80$ for $7$ out of $9$ test languages. The dataset and models are
available under open-source licences at
https://ai4bharat.iitm.ac.in/naamapadam. | cs.CL | Naamapadam: A Large-Scale Named Entity Annotated Data for Indic Languages |
2022-12-20T10:46:40Z | http://arxiv.org/pdf/2212.10154v2 | http://arxiv.org/abs/2212.10154v2 | Florian E. Dorner, Momchil Peychev, Nikola Konstantinov, Naman Goel, Elliott Ash, Martin Vechev | 2023-03-16T08:57:14Z | cs.CL, cs.AI, cs.CY, cs.LG | Text classifiers have promising applications in high-stake tasks such as
resume screening and content moderation. These classifiers must be fair and
avoid discriminatory decisions by being invariant to perturbations of sensitive
attributes such as gender or ethnicity. However, there is a gap between human
intuition about these perturbations and the formal similarity specifications
capturing them. While existing research has started to address this gap,
current methods are based on hardcoded word replacements, resulting in
specifications with limited expressivity or ones that fail to fully align with
human intuition (e.g., in cases of asymmetric counterfactuals). This work
proposes novel methods for bridging this gap by discovering expressive and
intuitive individual fairness specifications. We show how to leverage
unsupervised style transfer and GPT-3's zero-shot capabilities to automatically
generate expressive candidate pairs of semantically similar sentences that
differ along sensitive attributes. We then validate the generated pairs via an
extensive crowdsourcing study, which confirms that a lot of these pairs align
with human intuition about fairness in the context of toxicity classification.
Finally, we show how limited amounts of human feedback can be leveraged to
learn a similarity specification that can be used to train downstream
fairness-aware models. | cs.CL | Human-Guided Fair Classification for Natural Language Processing |
2022-12-20T10:44:18Z | http://arxiv.org/pdf/2212.10152v1 | http://arxiv.org/abs/2212.10152v1 | Risako Owan, Maria Gini, Dongyeop Kang | 2022-12-20T10:44:18Z | cs.CL | Modal verbs, such as "can", "may", and "must", are commonly used in daily
communication to convey the speaker's perspective related to the likelihood
and/or mode of the proposition. They can differ greatly in meaning depending on
how they're used and the context of a sentence (e.g. "They 'must' help each
other out." vs. "They 'must' have helped each other out.") Despite their
practical importance in natural language understanding, linguists have yet to
agree on a single, prominent framework for the categorization of modal verb
senses. This lack of agreement stems from high degrees of flexibility and
polysemy from the modal verbs, making it more difficult for researchers to
incorporate insights from this family of words into their work. This work
presents Moverb dataset, which consists of 27,240 annotations of modal verb
senses over 4,540 utterances containing one or more sentences from social
conversations. Each utterance is annotated by three annotators using two
different theoretical frameworks (i.e., Quirk and Palmer) of modal verb senses.
We observe that both frameworks have similar inter-annotator agreements,
despite having different numbers of sense types (8 for Quirk and 3 for Palmer).
With the RoBERTa-based classifiers fine-tuned on \dataset, we achieve F1 scores
of 82.2 and 78.3 on Quirk and Palmer, respectively, showing that modal verb
sense disambiguation is not a trivial task. Our dataset will be publicly
available with our final version. | cs.CL | Quirk or Palmer: A Comparative Study of Modal Verb Frameworks with Annotated Datasets |
2022-12-20T10:18:18Z | http://arxiv.org/pdf/2212.10140v2 | http://arxiv.org/abs/2212.10140v2 | Matthieu Futeral, Cordelia Schmid, Ivan Laptev, Benoît Sagot, Rachel Bawden | 2023-05-26T10:52:39Z | cs.CL, cs.CV | One of the major challenges of machine translation (MT) is ambiguity, which
can in some cases be resolved by accompanying context such as images. However,
recent work in multimodal MT (MMT) has shown that obtaining improvements from
images is challenging, limited not only by the difficulty of building effective
cross-modal representations, but also by the lack of specific evaluation and
training data. We present a new MMT approach based on a strong text-only MT
model, which uses neural adapters, a novel guided self-attention mechanism and
which is jointly trained on both visually-conditioned masking and MMT. We also
introduce CoMMuTE, a Contrastive Multilingual Multimodal Translation Evaluation
set of ambiguous sentences and their possible translations, accompanied by
disambiguating images corresponding to each translation. Our approach obtains
competitive results compared to strong text-only models on standard
English-to-French, English-to-German and English-to-Czech benchmarks and
outperforms baselines and state-of-the-art MMT systems by a large margin on our
contrastive test set. Our code and CoMMuTE are freely available. | cs.CL | Tackling Ambiguity with Images: Improved Multimodal Machine Translation and Contrastive Evaluation |
2022-12-20T09:34:43Z | http://arxiv.org/pdf/2212.10114v2 | http://arxiv.org/abs/2212.10114v2 | Maksym Del, Mark Fishel | 2023-06-01T18:50:21Z | cs.CL | Large language models (LLMs) have demonstrated solid zero-shot reasoning
capabilities, which is reflected in their performance on the current test
tasks. This calls for a more challenging benchmark requiring highly advanced
reasoning ability to be solved. In this paper, we introduce such a benchmark,
consisting of 191 long-form (1200 words on average) mystery narratives
constructed as detective puzzles. Puzzles are sourced from the "5 Minute
Mystery" platform and include a multiple-choice question for evaluation. Only
47% of humans solve a puzzle successfully on average, while the best human
solvers achieve over 80% success rate. We show that GPT-3 models barely
outperform random on this benchmark (with 28% accuracy) while state-of-the-art
GPT-4 solves only 38% of puzzles. This indicates that there is still a
significant gap in the deep reasoning abilities of LLMs and humans and
highlights the need for further research in this area. Our work introduces a
challenging benchmark for future studies on reasoning in language models and
contributes to a better understanding of the limits of LLMs' abilities. | cs.CL | True Detective: A Deep Abductive Reasoning Benchmark Undoable for GPT-3 and Challenging for GPT-4 |
2022-12-20T09:15:03Z | http://arxiv.org/pdf/2212.10097v2 | http://arxiv.org/abs/2212.10097v2 | Zhenyu Li, Xiuxing Li, Sunqi Fan, Jianyong Wang | 2024-06-21T03:06:36Z | cs.CL, cs.AI, cs.DB | Structured tabular data is a fundamental data type in numerous fields, and
the capacity to reason over tables is crucial for answering questions and
validating hypotheses. However, constructing labeled data for complex reasoning
tasks is labor intensive, and the quantity of annotated data remains
insufficient to support the intricate demands of real-world applications. To
address the insufficient annotation challenge, we present a self-training
framework for unsupervised complex tabular reasoning (UCTR-ST) by generating
diverse synthetic data with complex logic. Specifically, UCTR-ST incorporates
several essential techniques: we aggregate diverse programs and execute them on
tables based on a "Program-Management" component, and we bridge the gap between
programs and text with a powerful "Program-Transformation" module that
generates natural language sentences with complex logic. Furthermore, we
optimize the procedure using a "Table-Text Manipulator" to handle joint
table-text reasoning scenarios. The entire framework utilizes self-training
techniques to leverage the unlabeled training data, which results in
significant performance improvements when tested on real-world data.
Experimental results demonstrate that UCTRST achieves above 90% of the
supervised model performance on different tasks and domains, reducing the
dependence on manual annotation. Additionally, our approach can serve as a data
augmentation technique, significantly boosting the performance of supervised
models in low-resourced domains. | cs.CL | Optimization Techniques for Unsupervised Complex Table Reasoning via Self-Training Framework |
2022-12-20T08:55:47Z | http://arxiv.org/pdf/2212.10087v1 | http://arxiv.org/abs/2212.10087v1 | Yu Wang, Hongxia Jin | 2022-12-20T08:55:47Z | cs.CL | A coreference resolution system is to cluster all mentions that refer to the
same entity in a given context. All coreference resolution systems need to
tackle two main tasks: one task is to detect all of the potential mentions, and
the other is to learn the linking of an antecedent for each possible mention.
In this paper, we propose a hybrid rule-neural coreference resolution system
based on actor-critic learning, such that it can achieve better coreference
performance by leveraging the advantages from both the heuristic rules and a
neural conference model. This end-to-end system can also perform both mention
detection and resolution by leveraging a joint training algorithm. We
experiment on the BERT model to generate input span representations. Our model
with the BERT span representation achieves the state-of-the-art performance
among the models on the CoNLL-2012 Shared Task English Test Set. | cs.CL | Hybrid Rule-Neural Coreference Resolution System based on Actor-Critic Learning |
2022-12-20T08:43:10Z | http://arxiv.org/pdf/2212.10080v1 | http://arxiv.org/abs/2212.10080v1 | Shaswat Patel, Prince Bansal, Preeti Kaur | 2022-12-20T08:43:10Z | cs.CL, cs.LG | Recently, online social media has become a primary source for new information
and misinformation or rumours. In the absence of an automatic rumour detection
system the propagation of rumours has increased manifold leading to serious
societal damages. In this work, we propose a novel method for building
automatic rumour detection system by focusing on oversampling to alleviating
the fundamental challenges of class imbalance in rumour detection task. Our
oversampling method relies on contextualised data augmentation to generate
synthetic samples for underrepresented classes in the dataset. The key idea
exploits selection of tweets in a thread for augmentation which can be achieved
by introducing a non-random selection criteria to focus the augmentation
process on relevant tweets. Furthermore, we propose two graph neural
networks(GNN) to model non-linear conversations on a thread. To enhance the
tweet representations in our method we employed a custom feature selection
technique based on state-of-the-art BERTweet model. Experiments of three
publicly available datasets confirm that 1) our GNN models outperform the the
current state-of-the-art classifiers by more than 20%(F1-score); 2) our
oversampling technique increases the model performance by more than
9%;(F1-score) 3) focusing on relevant tweets for data augmentation via
non-random selection criteria can further improve the results; and 4) our
method has superior capabilities to detect rumours at very early stage. | cs.CL | Rumour detection using graph neural network and oversampling in benchmark Twitter dataset |
2022-12-20T08:34:56Z | http://arxiv.org/pdf/2212.10079v1 | http://arxiv.org/abs/2212.10079v1 | Yichen Xu, Yanqiao Zhu | 2022-12-20T08:34:56Z | cs.SE, cs.CL, cs.LG | As the complexity of modern software continues to escalate, software
engineering has become an increasingly daunting and error-prone endeavor. In
recent years, the field of Neural Code Intelligence (NCI) has emerged as a
promising solution, leveraging the power of deep learning techniques to tackle
analytical tasks on source code with the goal of improving programming
efficiency and minimizing human errors within the software industry. Pretrained
language models have become a dominant force in NCI research, consistently
delivering state-of-the-art results across a wide range of tasks, including
code summarization, generation, and translation. In this paper, we present a
comprehensive survey of the NCI domain, including a thorough review of
pretraining techniques, tasks, datasets, and model architectures. We hope this
paper will serve as a bridge between the natural language and programming
language communities, offering insights for future research in this rapidly
evolving field. | cs.SE | A Survey on Pretrained Language Models for Neural Code Intelligence |
2022-12-20T08:30:58Z | http://arxiv.org/pdf/2212.10077v3 | http://arxiv.org/abs/2212.10077v3 | Kevin Yang, Dan Klein, Nanyun Peng, Yuandong Tian | 2023-06-14T22:05:11Z | cs.CL, cs.AI | We propose the Detailed Outline Control (DOC) framework for improving
long-range plot coherence when automatically generating
several-thousand-word-long stories. DOC consists of two complementary
components: a detailed outliner and a detailed controller. The detailed
outliner creates a more detailed, hierarchically structured outline, shifting
creative burden from the main drafting procedure to the planning stage. The
detailed controller ensures the more detailed outline is still respected during
generation by controlling story passages to align with outline details. In
human evaluations of automatically generated stories, DOC substantially
outperforms a strong Re3 baseline (Yang et al., 2022) on plot coherence (22.5%
absolute gain), outline relevance (28.2%), and interestingness (20.7%). Humans
also judged DOC to be much more controllable in an interactive generation
setting. | cs.CL | DOC: Improving Long Story Coherence With Detailed Outline Control |
2022-12-20T08:24:45Z | http://arxiv.org/pdf/2212.10071v2 | http://arxiv.org/abs/2212.10071v2 | Namgyu Ho, Laura Schmid, Se-Young Yun | 2023-06-13T10:55:29Z | cs.CL, cs.AI, cs.LG | Recent works have shown that chain-of-thought (CoT) prompting can elicit
language models to solve complex reasoning tasks, step-by-step. However,
prompt-based CoT methods are dependent on very large models such as GPT-3 175B
which are prohibitive to deploy at scale. In this paper, we use these large
models as reasoning teachers to enable complex reasoning in smaller models and
reduce model size requirements by several orders of magnitude. We propose
Fine-tune-CoT, a method that generates reasoning samples from very large
teacher models to fine-tune smaller models. We evaluate our method on a wide
range of public models and complex tasks. We find that Fine-tune-CoT enables
substantial reasoning capability in small models, far outperforming
prompt-based baselines and even the teacher model in many tasks. Additionally,
we extend our method by leveraging the teacher model's ability to generate
multiple distinct rationales for each original sample. Enriching the
fine-tuning data with such diverse reasoning results in a substantial
performance boost across datasets, even for very small models. We conduct
ablations and sample studies to understand the emergence of reasoning
capabilities of student models. Our code implementation and data are available
at https://github.com/itsnamgyu/reasoning-teacher. | cs.CL | Large Language Models Are Reasoning Teachers |
2022-12-20T08:06:55Z | http://arxiv.org/pdf/2212.10060v2 | http://arxiv.org/abs/2212.10060v2 | Pei Zhou, Andrew Zhu, Jennifer Hu, Jay Pujara, Xiang Ren, Chris Callison-Burch, Yejin Choi, Prithviraj Ammanabrolu | 2023-05-30T23:09:37Z | cs.CL, cs.AI | We propose a novel task, G4C, to study teacher-student natural language
interactions in a goal-driven and grounded environment. Dungeons and Dragons
(D&D), a role-playing game, provides an ideal setting to investigate such
interactions. Here, the Dungeon Master (DM), i.e., the teacher, guides the
actions of several players -- students, each with their own personas and
abilities -- to achieve shared goals grounded in a fantasy world. Our approach
is to decompose and model these interactions into (1) the DM's intent to guide
players toward a given goal; (2) the DM's guidance utterance to the players
expressing this intent; and (3) a theory-of-mind (ToM) model that anticipates
the players' reaction to the guidance one turn into the future. We develop a
novel reinforcement learning (RL) method for training a DM that generates
guidance for players by rewarding utterances where the intent matches the
ToM-anticipated player actions. Human and automated evaluations show that a DM
trained to explicitly model intents and incorporate ToM of the players using RL
generates better-quality guidance that is 3x more likely to fulfill the DM's
intent than a vanilla natural language generation (NLG) approach. | cs.CL | I Cast Detect Thoughts: Learning to Converse and Guide with Intents and Theory-of-Mind in Dungeons and Dragons |
2022-12-20T08:04:36Z | http://arxiv.org/pdf/2212.10057v2 | http://arxiv.org/abs/2212.10057v2 | Wenhao Wu, Wei Li, Xinyan Xiao, Jiachen Liu, Sujian Li, Yajuan Lv | 2023-05-27T13:42:23Z | cs.CL | A crucial issue of current text generation models is that they often
uncontrollably generate factually inconsistent text with respective of their
inputs. Limited by the lack of annotated data, existing works in evaluating
factual consistency directly transfer the reasoning ability of models trained
on other data-rich upstream tasks like question answering (QA) and natural
language inference (NLI) without any further adaptation. As a result, they
perform poorly on the real generated text and are biased heavily by their
single-source upstream tasks. To alleviate this problem, we propose a weakly
supervised framework that aggregates multiple resources to train a precise and
efficient factual metric, namely WeCheck. WeCheck first utilizes a generative
model to accurately label a real generated sample by aggregating its weak
labels, which are inferred from multiple resources. Then, we train the target
metric model with the weak supervision while taking noises into consideration.
Comprehensive experiments on a variety of tasks demonstrate the strong
performance of WeCheck, which achieves a 3.4\% absolute improvement over
previous state-of-the-art methods on TRUE benchmark on average. | cs.CL | WeCheck: Strong Factual Consistency Checker via Weakly Supervised Learning |
2022-12-20T07:54:58Z | http://arxiv.org/pdf/2212.10051v1 | http://arxiv.org/abs/2212.10051v1 | Subhasis Dasgupta, Jaydip Sen | 2022-12-20T07:54:58Z | cs.CL, cs.LG | Opinion mining is the branch of computation that deals with opinions,
appraisals, attitudes, and emotions of people and their different aspects. This
field has attracted substantial research interest in recent years. Aspect-level
(called aspect-based opinion mining) is often desired in practical applications
as it provides detailed opinions or sentiments about different aspects of
entities and entities themselves, which are usually required for action. Aspect
extraction and entity extraction are thus two core tasks of aspect-based
opinion mining. his paper has presented a framework of aspect-based opinion
mining based on the concept of transfer learning. on real-world customer
reviews available on the Amazon website. The model has yielded quite
satisfactory results in its task of aspect-based opinion mining. | cs.CL | A Framework of Customer Review Analysis Using the Aspect-Based Opinion Mining Approach |
2022-12-20T07:44:25Z | http://arxiv.org/pdf/2212.10047v2 | http://arxiv.org/abs/2212.10047v2 | Jing Xie, James B. Wendt, Yichao Zhou, Seth Ebner, Sandeep Tata | 2022-12-22T09:13:04Z | cs.CL | Many business workflows require extracting important fields from form-like
documents (e.g. bank statements, bills of lading, purchase orders, etc.).
Recent techniques for automating this task work well only when trained with
large datasets. In this work we propose a novel data augmentation technique to
improve performance when training data is scarce, e.g. 10-250 documents. Our
technique, which we call FieldSwap, works by swapping out the key phrases of a
source field with the key phrases of a target field to generate new synthetic
examples of the target field for use in training. We demonstrate that this
approach can yield 1-7 F1 point improvements in extraction performance. | cs.CL | An Augmentation Strategy for Visually Rich Documents |
2022-12-20T07:22:45Z | http://arxiv.org/pdf/2212.10039v1 | http://arxiv.org/abs/2212.10039v1 | Tanmay Chavan, Shantanu Patankar, Aditya Kane, Omkar Gokhale, Raviraj Joshi | 2022-12-20T07:22:45Z | cs.CL | Automated offensive language detection is essential in combating the spread
of hate speech, particularly in social media. This paper describes our work on
Offensive Language Identification in low resource Indic language Marathi. The
problem is formulated as a text classification task to identify a tweet as
offensive or non-offensive. We evaluate different mono-lingual and
multi-lingual BERT models on this classification task, focusing on BERT models
pre-trained with social media datasets. We compare the performance of MuRIL,
MahaTweetBERT, MahaTweetBERT-Hateful, and MahaBERT on the HASOC 2022 test set.
We also explore external data augmentation from other existing Marathi hate
speech corpus HASOC 2021 and L3Cube-MahaHate. The MahaTweetBERT, a BERT model,
pre-trained on Marathi tweets when fine-tuned on the combined dataset (HASOC
2021 + HASOC 2022 + MahaHate), outperforms all models with an F1 score of 98.43
on the HASOC 2022 test set. With this, we also provide a new state-of-the-art
result on HASOC 2022 / MOLD v2 test set. | cs.CL | A Twitter BERT Approach for Offensive Language Detection in Marathi |
2022-12-20T06:54:04Z | http://arxiv.org/pdf/2212.10029v3 | http://arxiv.org/abs/2212.10029v3 | Yuling Gu, Bhavana Dalvi Mishra, Peter Clark | 2023-06-08T17:27:44Z | cs.CL, cs.AI | When people think of everyday things like an egg, they typically have a
mental image associated with it. This allows them to correctly judge, for
example, that "the yolk surrounds the shell" is a false statement. Do language
models similarly have a coherent picture of such everyday things? To
investigate this, we propose a benchmark dataset consisting of 100 everyday
things, their parts, and the relationships between these parts, expressed as
11,720 "X relation Y?" true/false questions. Using these questions as probes,
we observe that state-of-the-art pre-trained language models (LMs) like GPT-3
and Macaw have fragments of knowledge about these everyday things, but do not
have fully coherent "parts mental models" (54-59% accurate, 19-43% conditional
constraint violation). We propose an extension where we add a constraint
satisfaction layer on top of the LM's raw predictions to apply commonsense
constraints. As well as removing inconsistencies, we find that this also
significantly improves accuracy (by 16-20%), suggesting how the incoherence of
the LM's pictures of everyday things can be significantly reduced. | cs.CL | Do language models have coherent mental models of everyday things? |
2022-12-20T06:44:32Z | http://arxiv.org/pdf/2212.10025v2 | http://arxiv.org/abs/2212.10025v2 | Zhuo Zhang, Yuanhang Yang, Yong Dai, Lizhen Qu, Zenglin Xu | 2023-06-02T12:58:07Z | cs.LG, cs.CL | With increasing privacy concerns on data, recent studies have made
significant progress using federated learning (FL) on privacy-sensitive natural
language processing (NLP) tasks. Much literature suggests fully fine-tuning
pre-trained language models (PLMs) in the FL paradigm can mitigate the data
heterogeneity problem and close the performance gap with centralized training.
However, large PLMs bring the curse of prohibitive communication overhead and
local model adaptation costs for the FL system. To this end, we introduce
various parameter-efficient tuning (PETuning) methods into federated learning.
Specifically, we provide a holistic empirical study of representative PLMs
tuning methods in FL. The experimental results cover the analysis of data
heterogeneity levels, data scales, and different FL scenarios. Overall
communication overhead can be significantly reduced by locally tuning and
globally aggregating lightweight model parameters while maintaining acceptable
performance in various FL settings. To facilitate the research of PETuning in
FL, we also develop a federated tuning framework FedPETuning, which allows
practitioners to exploit different PETuning methods under the FL training
paradigm conveniently. The source code is available at
\url{https://github.com/iezhuozhuo/FedETuning/tree/deltaTuning}. | cs.LG | When Federated Learning Meets Pre-trained Language Models' Parameter-Efficient Tuning Methods |
2022-12-20T06:24:25Z | http://arxiv.org/pdf/2212.10020v3 | http://arxiv.org/abs/2212.10020v3 | Tianxing He, Jingyu Zhang, Tianle Wang, Sachin Kumar, Kyunghyun Cho, James Glass, Yulia Tsvetkov | 2023-05-18T20:46:57Z | cs.CL | In this work, we explore a useful but often neglected methodology for
robustness analysis of text generation evaluation metrics: stress tests with
synthetic data. Basically, we design and synthesize a wide range of potential
errors and check whether they result in a commensurate drop in the metric
scores. We examine a range of recently proposed evaluation metrics based on
pretrained language models, for the tasks of open-ended generation,
translation, and summarization. Our experiments reveal interesting
insensitivities, biases, or even loopholes in existing metrics. For example, we
find that BERTScore is confused by truncation errors in summarization, and
MAUVE (built on top of GPT-2) is insensitive to errors at the beginning or
middle of generations. Further, we investigate the reasons behind these blind
spots and suggest practical workarounds for a more reliable evaluation of text
generation. We have released our code and data at
https://github.com/cloudygoose/blindspot_nlg. | cs.CL | On the Blind Spots of Model-Based Evaluation Metrics for Text Generation |
2022-12-20T06:23:02Z | http://arxiv.org/pdf/2212.10019v1 | http://arxiv.org/abs/2212.10019v1 | Kangda Wei, Dawn Lawrie, Benjamin Van Durme, Yunmo Chen, Orion Weller | 2022-12-20T06:23:02Z | cs.CL | Answering complex questions often requires multi-step reasoning in order to
obtain the final answer. Most research into decompositions of complex questions
involves open-domain systems, which have shown success in using these
decompositions for improved retrieval. In the machine reading setting, however,
work to understand when decompositions are helpful is understudied. We conduct
experiments on decompositions in machine reading to unify recent work in this
space, using a range of models and datasets. We find that decompositions can be
helpful in the few-shot case, giving several points of improvement in exact
match scores. However, we also show that when models are given access to
datasets with around a few hundred or more examples, decompositions are not
helpful (and can actually be detrimental). Thus, our analysis implies that
models can learn decompositions implicitly even with limited data. | cs.CL | When Do Decompositions Help for Machine Reading? |
2022-12-20T06:21:21Z | http://arxiv.org/pdf/2212.10018v2 | http://arxiv.org/abs/2212.10018v2 | Yu Li, Baolin Peng, Pengcheng He, Michel Galley, Zhou Yu, Jianfeng Gao | 2023-05-26T17:29:01Z | cs.CL | Dialogue summarization has recently garnered significant attention due to its
wide range of applications. However, existing methods for summarizing dialogues
have limitations because they do not take into account the inherent structure
of dialogue and rely heavily on labeled data, which can lead to poor
performance in new domains. In this work, we propose DIONYSUS (dynamic input
optimization in pre-training for dialogue summarization), a pre-trained
encoder-decoder model for summarizing dialogues in any new domain. To pre-train
DIONYSUS, we create two pseudo summaries for each dialogue example: one is
produced by a fine-tuned summarization model, and the other is a collection of
dialogue turns that convey important information. We then choose one of these
pseudo summaries based on the difference in information distribution across
different types of dialogues. This selected pseudo summary serves as the
objective for pre-training DIONYSUS using a self-supervised approach on a large
dialogue corpus. Our experiments show that DIONYSUS outperforms existing
methods on six datasets, as demonstrated by its ROUGE scores in zero-shot and
few-shot settings. | cs.CL | DIONYSUS: A Pre-trained Model for Low-Resource Dialogue Summarization |
2022-12-20T06:03:51Z | http://arxiv.org/pdf/2212.10015v3 | http://arxiv.org/abs/2212.10015v3 | Tejas Gokhale, Hamid Palangi, Besmira Nushi, Vibhav Vineet, Eric Horvitz, Ece Kamar, Chitta Baral, Yezhou Yang | 2023-10-27T17:24:04Z | cs.CV, cs.AI, cs.CL | Spatial understanding is a fundamental aspect of computer vision and integral
for human-level reasoning about images, making it an important component for
grounded language understanding. While recent text-to-image synthesis (T2I)
models have shown unprecedented improvements in photorealism, it is unclear
whether they have reliable spatial understanding capabilities. We investigate
the ability of T2I models to generate correct spatial relationships among
objects and present VISOR, an evaluation metric that captures how accurately
the spatial relationship described in text is generated in the image. To
benchmark existing models, we introduce a dataset, $\mathrm{SR}_{2D}$, that
contains sentences describing two or more objects and the spatial relationships
between them. We construct an automated evaluation pipeline to recognize
objects and their spatial relationships, and employ it in a large-scale
evaluation of T2I models. Our experiments reveal a surprising finding that,
although state-of-the-art T2I models exhibit high image quality, they are
severely limited in their ability to generate multiple objects or the specified
spatial relations between them. Our analyses demonstrate several biases and
artifacts of T2I models such as the difficulty with generating multiple
objects, a bias towards generating the first object mentioned, spatially
inconsistent outputs for equivalent relationships, and a correlation between
object co-occurrence and spatial understanding capabilities. We conduct a human
study that shows the alignment between VISOR and human judgement about spatial
understanding. We offer the $\mathrm{SR}_{2D}$ dataset and the VISOR metric to
the community in support of T2I reasoning research. | cs.CV | Benchmarking Spatial Relationships in Text-to-Image Generation |
2022-12-20T06:01:13Z | http://arxiv.org/pdf/2212.10013v2 | http://arxiv.org/abs/2212.10013v2 | Forrest Sheng Bao, Ruixuan Tu, Ge Luo, Yinfei Yang, Hebi Li, Minghui Qiu, Youbiao He, Cen Chen | 2023-11-26T05:05:51Z | cs.AI, cs.CL | Automated summary quality assessment falls into two categories:
reference-based and reference-free. Reference-based metrics, historically
deemed more accurate due to the additional information provided by
human-written references, are limited by their reliance on human input. In this
paper, we hypothesize that the comparison methodologies used by some
reference-based metrics to evaluate a system summary against its corresponding
reference can be effectively adapted to assess it against its source document,
thereby transforming these metrics into reference-free ones. Experimental
results support this hypothesis. After being repurposed reference-freely, the
zero-shot BERTScore using the pretrained DeBERTa-large-MNLI model of <0.5B
parameters consistently outperforms its original reference-based version across
various aspects on the SummEval and Newsroom datasets. It also excels in
comparison to most existing reference-free metrics and closely competes with
zero-shot summary evaluators based on GPT-3.5. | cs.AI | DocAsRef: An Empirical Study on Repurposing Reference-Based Summary Quality Metrics Reference-Freely |
2022-12-20T05:59:42Z | http://arxiv.org/pdf/2212.10012v1 | http://arxiv.org/abs/2212.10012v1 | Belinda Z. Li, Maxwell Nye, Jacob Andreas | 2022-12-20T05:59:42Z | cs.CL | Language models (LMs) often generate incoherent outputs: they refer to events
and entity states that are incompatible with the state of the world described
in their inputs. We introduce SituationSupervision, a family of approaches for
improving coherence in LMs by training them to construct and condition on
explicit representations of entities and their states. SituationSupervision has
two components: an auxiliary situation modeling task that trains models to
predict state representations in context, and a latent state inference
procedure that imputes these states from partially annotated training data.
SituationSupervision can be applied to both fine-tuning (by supervising LMs to
encode state variables in their hidden representations) and prompting (by
inducing LMs to interleave textual descriptions of entity states with output
text). In both cases, SituationSupervision requires only a small number of
state annotations to produce major coherence improvements (between 4-11%),
showing that standard LMs can be sample-efficiently trained to model not just
language but the situations it describes. | cs.CL | Language Modeling with Latent Situations |