Word in Context (WIC)

Original Paper: https://wic-ita.github.io/

This dataset comes from EVALITA-2023.

Word in Context task consists of establishing if a word w occurring in two different sentences s1 and s2 has the same meaning or not.

We repropose this task to test generative LLMs defining a specific prompting strategy comparing the perplexities of possible continuations to understand the models' capabilities.

Example

Here you can see the structure of the single sample in the present dataset.

{
  "sentence_1": string, # text of the sentence 1
  "sentence_2": string, # text of the sentence 2
  "lemma": string, # text of the word present in both sentences
  "label": int, # 0: Different Mearning, 1: Same Meaning,
}

Statistics

WIC	0	1
Training	806	1999
Validation	250	250
Test	250	250

Proposed Prompts

Here we will describe the prompt given to the model over which we will compute the perplexity score, as model's answer we will chose the prompt with lower perplexity. Moreover, for each subtask, we define a description that is prepended to the prompts, needed by the model to understand the task.

Description of the task: "Date due frasi, che contengono un lemma in comune, indica se tale lemma ha lo stesso significato in entrambe le frasi.\n\n"

Cloze Style:

Label (Different Meaning): "Frase 1: {{sentence1}}\nFrase 2: {{sentence2}}\nLa parola '{{lemma}}' nelle due frasi precedenti ha un significato differente tra le due frasi"

Label (Same Meaning): "Frase 1: {{sentence1}}\nFrase 2: {{sentence2}}\nLa parola '{{lemma}}' nelle due frasi precedenti ha lo stesso significato in entrambe le frasi"

MCQA Style:

Frase 1: {{sentence1}}\nFrase 2: {{sentence2}}\nDomanda: La parola \"{{lemma}}\" ha lo stesso signicato nelle due frasi precedenti? Rispondi sì o no:

Results

The following results are given by the Cloze-style prompting over some english and italian-adapted LLMs.

WIC	ACCURACY (5-shots)
Gemma-2B	48.2
QWEN2-1.5B	50.4
Mistral-7B	53.4
ZEFIRO	54.6
Llama-3-8B	54.6
Llama-3-8B-IT	62.8
ANITA	69.2

Acknowledge

We would like to thank the authors of this resource for publicly releasing such an intriguing benchmark.

Additionally, we extend our gratitude to the students of the MNLP-2024 course, whose first homework explored various interesting prompting strategies.

The original dataset is freely available for download link.

License

Original data license not found.