metadata

license: cc-by-4.0
metrics:
  - bleu4
  - meteor
  - rouge-l
  - bertscore
  - moverscore
language: ru
datasets:
  - lmqg/qag_ruquad
pipeline_tag: text2text-generation
tags:
  - questions and answers generation
widget:
  - text: >-
      Нелишним будет отметить, что, развивая это направление, Д. И. Менделеев,
      поначалу априорно выдвинув идею о температуре, при которой высота мениска
      будет нулевой, в мае 1860 года провёл серию опытов.
    example_title: Questions & Answers Generation Example 1
model-index:
  - name: lmqg/mbart-large-cc25-ruquad-qag
    results:
      - task:
          name: Text2text Generation
          type: text2text-generation
        dataset:
          name: lmqg/qag_ruquad
          type: default
          args: default
        metrics:
          - name: BLEU4 (Question & Answer Generation)
            type: bleu4_question_answer_generation
            value: 4.33
          - name: ROUGE-L (Question & Answer Generation)
            type: rouge_l_question_answer_generation
            value: 18.59
          - name: METEOR (Question & Answer Generation)
            type: meteor_question_answer_generation
            value: 23.52
          - name: BERTScore (Question & Answer Generation)
            type: bertscore_question_answer_generation
            value: 69.58
          - name: MoverScore (Question & Answer Generation)
            type: moverscore_question_answer_generation
            value: 52.29
          - name: QAAlignedF1Score-BERTScore (Question & Answer Generation)
            type: qa_aligned_f1_score_bertscore_question_answer_generation
            value: 77.36
          - name: QAAlignedRecall-BERTScore (Question & Answer Generation)
            type: qa_aligned_recall_bertscore_question_answer_generation
            value: 80.05
          - name: QAAlignedPrecision-BERTScore (Question & Answer Generation)
            type: qa_aligned_precision_bertscore_question_answer_generation
            value: 74.97
          - name: QAAlignedF1Score-MoverScore (Question & Answer Generation)
            type: qa_aligned_f1_score_moverscore_question_answer_generation
            value: 56.1
          - name: QAAlignedRecall-MoverScore (Question & Answer Generation)
            type: qa_aligned_recall_moverscore_question_answer_generation
            value: 58.11
          - name: QAAlignedPrecision-MoverScore (Question & Answer Generation)
            type: qa_aligned_precision_moverscore_question_answer_generation
            value: 54.4

Model Card of `lmqg/mbart-large-cc25-ruquad-qag`

This model is fine-tuned version of facebook/mbart-large-cc25 for question & answer pair generation task on the lmqg/qag_ruquad (dataset_name: default) via lmqg.

Overview

Language model: facebook/mbart-large-cc25
Language: ru
Training data: lmqg/qag_ruquad (default)
Online Demo: https://autoqg.net/
Repository: https://github.com/asahi417/lm-question-generation
Paper: https://arxiv.org/abs/2210.03992

Usage

With lmqg

from lmqg import TransformersQG

# initialize model
model = TransformersQG(language="ru", model="lmqg/mbart-large-cc25-ruquad-qag")

# model prediction
question_answer_pairs = model.generate_qa("Нелишним будет отметить, что, развивая это направление, Д. И. Менделеев, поначалу априорно выдвинув идею о температуре, при которой высота мениска будет нулевой, в мае 1860 года провёл серию опытов.")

With transformers

from transformers import pipeline

pipe = pipeline("text2text-generation", "lmqg/mbart-large-cc25-ruquad-qag")
output = pipe("Нелишним будет отметить, что, развивая это направление, Д. И. Менделеев, поначалу априорно выдвинув идею о температуре, при которой высота мениска будет нулевой, в мае 1860 года провёл серию опытов.")

Evaluation

Metric (Question & Answer Generation): raw metric file

	Score	Type	Dataset
BERTScore	69.58	default	lmqg/qag_ruquad
Bleu_1	13.13	default	lmqg/qag_ruquad
Bleu_2	8.29	default	lmqg/qag_ruquad
Bleu_3	5.85	default	lmqg/qag_ruquad
Bleu_4	4.33	default	lmqg/qag_ruquad
METEOR	23.52	default	lmqg/qag_ruquad
MoverScore	52.29	default	lmqg/qag_ruquad
QAAlignedF1Score (BERTScore)	77.36	default	lmqg/qag_ruquad
QAAlignedF1Score (MoverScore)	56.1	default	lmqg/qag_ruquad
QAAlignedPrecision (BERTScore)	74.97	default	lmqg/qag_ruquad
QAAlignedPrecision (MoverScore)	54.4	default	lmqg/qag_ruquad
QAAlignedRecall (BERTScore)	80.05	default	lmqg/qag_ruquad
QAAlignedRecall (MoverScore)	58.11	default	lmqg/qag_ruquad
ROUGE_L	18.59	default	lmqg/qag_ruquad

Training hyperparameters

The following hyperparameters were used during fine-tuning:

dataset_path: lmqg/qag_ruquad
dataset_name: default
input_types: ['paragraph']
output_types: ['questions_answers']
prefix_types: None
model: facebook/mbart-large-cc25
max_length: 512
max_length_output: 256
epoch: 6
batch: 2
lr: 0.0001
fp16: False
random_seed: 1
gradient_accumulation_steps: 32
label_smoothing: 0.15

The full configuration can be found at fine-tuning config file.

Citation

@inproceedings{ushio-etal-2022-generative,
    title = "{G}enerative {L}anguage {M}odels for {P}aragraph-{L}evel {Q}uestion {G}eneration",
    author = "Ushio, Asahi  and
        Alva-Manchego, Fernando  and
        Camacho-Collados, Jose",
    booktitle = "Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing",
    month = dec,
    year = "2022",
    address = "Abu Dhabi, U.A.E.",
    publisher = "Association for Computational Linguistics",
}