| """SemanticQA: A Semantic Reasoning Benchmark for Language Models.""" |
|
|
| import json |
| import os |
|
|
| import datasets |
|
|
|
|
| _CITATION = """\ |
| @article{liu2024revisiting, |
| title={Revisiting a Pain in the Neck: Semantic Phrase Processing Benchmark for Language Models}, |
| author={Liu, Yang and Qin, Melissa Xiaohui and Li, Hongming and Huang, Chao}, |
| journal={arXiv preprint arXiv:2405.02861}, |
| year={2024} |
| } |
| """ |
|
|
| _DESCRIPTION = """\ |
| SemanticQA is a comprehensive benchmark for evaluating language models on semantic \ |
| phrase processing tasks, covering idioms, noun compounds, lexical collocations, and \ |
| verbal multiword expressions (VMWEs). It includes 11 core evaluation subsets spanning \ |
| 4 phrase types with tasks such as detection, extraction, categorization, interpretation, \ |
| and retrieval. |
| """ |
|
|
| _HOMEPAGE = "https://github.com/jacklanda/SemanticQA" |
| _LICENSE = "MIT" |
| _DATA_DIR = "data" |
|
|
| _CONFIGS = { |
| "collocate_retrieval": { |
| "description": "Collocate Retrieval (CR): Given a base word, its lexical function label, and a sentential context with a [MASK] token, retrieve the correct collocate.", |
| "data_files": {"test": "collocate_retrieval/collocate_retrieval.json"}, |
| "features": datasets.Features({ |
| "id": datasets.Value("string"), |
| "base": datasets.Value("string"), |
| "collocate": datasets.Value("string"), |
| "collocation": datasets.Value("string"), |
| "label": datasets.Value("string"), |
| "context": datasets.Value("string"), |
| }), |
| }, |
| "collocation_categorization": { |
| "description": "Lexical Collocation Categorization (LCC): Classify a collocation into its lexical function category.", |
| "data_files": {"test": "collocation_categorization/collocation_categorization.json"}, |
| "features": datasets.Features({ |
| "id": datasets.Value("string"), |
| "base": datasets.Value("string"), |
| "collocate": datasets.Value("string"), |
| "collocation": datasets.Value("string"), |
| "label": datasets.Value("string"), |
| "label_id": datasets.Value("string"), |
| "context": datasets.Value("string"), |
| }), |
| }, |
| "collocation_extraction": { |
| "description": "Lexical Collocation Extraction (LCE): Extract the collocation from a given context.", |
| "data_files": {"test": "collocation_extraction/collocation_extraction.json"}, |
| "features": datasets.Features({ |
| "id": datasets.Value("string"), |
| "keyword": datasets.Value("string"), |
| "value": datasets.Value("string"), |
| "collocation": datasets.Value("string"), |
| "label": datasets.Value("string"), |
| "label_id": datasets.Value("string"), |
| "context": datasets.Value("string"), |
| }), |
| }, |
| "collocation_paraphrase": { |
| "description": "Lexical Collocation Interpretation (LCI): Generate paraphrases for a collocation in context.", |
| "data_files": {"test": "collocation_paraphrase/collocation_paraphrase.json"}, |
| "features": datasets.Features({ |
| "id": datasets.Value("string"), |
| "base": datasets.Value("string"), |
| "collocate": datasets.Value("string"), |
| "collocation": datasets.Value("string"), |
| "label": datasets.Value("string"), |
| "label_id": datasets.Value("string"), |
| "context": datasets.Value("string"), |
| "paraphrases": datasets.Sequence(datasets.Value("string")), |
| }), |
| }, |
| "idiom_detection": { |
| "description": "Idiomatic Expression Detection (IED): Multiple-choice task to identify the meaning of an idiom in context.", |
| "data_files": {"test": "idiom_detection/idiom_detection.json"}, |
| "features": datasets.Features({ |
| "id": datasets.Value("string"), |
| "context": datasets.Value("string"), |
| "idiom": datasets.Value("string"), |
| "A": datasets.Value("string"), |
| "B": datasets.Value("string"), |
| "C": datasets.Value("string"), |
| "D": datasets.Value("string"), |
| "target": datasets.Value("string"), |
| }), |
| }, |
| "idiom_extraction": { |
| "description": "Idiomatic Expression Extraction (IEE): Extract the idiomatic expression from a given context.", |
| "data_files": {"test": "idiom_extraction/idiom_extraction.json"}, |
| "features": datasets.Features({ |
| "context": datasets.Value("string"), |
| "idiom": datasets.Value("string"), |
| }), |
| }, |
| "idiom_paraphrase": { |
| "description": "Idiomatic Expression Interpretation (IEI): Generate a literal paraphrase of an idiomatic expression in context.", |
| "data_files": {"test": "idiom_paraphrase/idiom_paraphrase.json"}, |
| "features": datasets.Features({ |
| "id": datasets.Value("string"), |
| "idiom": datasets.Value("string"), |
| "paraphrase": datasets.Value("string"), |
| "context_idiomatic": datasets.Value("string"), |
| "context_literal": datasets.Value("string"), |
| }), |
| }, |
| "noun_compound_compositionality": { |
| "description": "Noun Compound Compositionality (NCC): Multiple-choice task to judge the compositionality level of a noun compound in context.", |
| "data_files": {"test": "noun_compound_compositionality/noun_compound_compositionality.json"}, |
| "features": datasets.Features({ |
| "id": datasets.Value("string"), |
| "noun_compound": datasets.Value("string"), |
| "context": datasets.Value("string"), |
| "A": datasets.Value("string"), |
| "B": datasets.Value("string"), |
| "C": datasets.Value("string"), |
| "D": datasets.Value("string"), |
| "target": datasets.Value("string"), |
| }), |
| }, |
| "noun_compound_compositionality_ft": { |
| "description": "Noun Compound Compositionality fine-tuning splits (NCC-FT): Train/test/validation data for fine-tuning compositionality classifiers.", |
| "data_files": { |
| "train": "noun_compound_compositionality/noun_compound_compositionality_ft_train.json", |
| "test": "noun_compound_compositionality/noun_compound_compositionality_ft_test.json", |
| "validation": "noun_compound_compositionality/noun_compound_compositionality_ft_valid.json", |
| }, |
| "features": datasets.Features({ |
| "text": datasets.Value("string"), |
| "options": datasets.Value("string"), |
| "answer": datasets.Value("string"), |
| }), |
| }, |
| "noun_compound_extraction": { |
| "description": "Noun Compound Extraction (NCE): Extract the noun compound from a given context.", |
| "data_files": {"test": "noun_compound_extraction/noun_compound_extraction.json"}, |
| "features": datasets.Features({ |
| "context": datasets.Value("string"), |
| "start_index": datasets.Value("string"), |
| "end_index": datasets.Value("string"), |
| "noun_compound": datasets.Value("string"), |
| "interpretation": datasets.Value("string"), |
| }), |
| }, |
| "noun_compound_interpretation": { |
| "description": "Noun Compound Interpretation (NCI): Generate free-form interpretations of a noun compound.", |
| "data_files": {"test": "noun_compound_interpretation/noun_compound_interpretation.json"}, |
| "features": datasets.Features({ |
| "id": datasets.Value("string"), |
| "noun_compound": datasets.Value("string"), |
| "references": datasets.Sequence(datasets.Value("string")), |
| }), |
| }, |
| "verbal_mwe_extraction": { |
| "description": "Verbal Multiword Expression Extraction (VMWE): Extract the verbal MWE from a given context and identify its type (VPC, LVC, VID).", |
| "data_files": {"test": "verbal_mwe_extraction/verbal_mwe_extraction.json"}, |
| "features": datasets.Features({ |
| "id": datasets.Value("string"), |
| "context": datasets.Value("string"), |
| "vmwe": datasets.Value("string"), |
| "label": datasets.Value("string"), |
| }), |
| }, |
| } |
|
|
|
|
| class SemanticQA(datasets.GeneratorBasedBuilder): |
| """SemanticQA benchmark dataset.""" |
|
|
| BUILDER_CONFIGS = [ |
| datasets.BuilderConfig(name=name, description=cfg["description"]) |
| for name, cfg in _CONFIGS.items() |
| ] |
| DEFAULT_CONFIG_NAME = "idiom_detection" |
|
|
| def _info(self): |
| cfg = _CONFIGS[self.config.name] |
| return datasets.DatasetInfo( |
| description=cfg["description"], |
| features=cfg["features"], |
| homepage=_HOMEPAGE, |
| license=_LICENSE, |
| citation=_CITATION, |
| ) |
|
|
| def _split_generators(self, dl_manager): |
| cfg = _CONFIGS[self.config.name] |
| urls = { |
| split: os.path.join(_DATA_DIR, path) |
| for split, path in cfg["data_files"].items() |
| } |
| downloaded = dl_manager.download_and_extract(urls) |
|
|
| split_map = { |
| "test": datasets.Split.TEST, |
| "train": datasets.Split.TRAIN, |
| "validation": datasets.Split.VALIDATION, |
| } |
| return [ |
| datasets.SplitGenerator( |
| name=split_map[split_key], |
| gen_kwargs={"filepath": path}, |
| ) |
| for split_key, path in downloaded.items() |
| ] |
|
|
| def _generate_examples(self, filepath): |
| with open(filepath, encoding="utf-8") as f: |
| data = json.load(f) |
| for idx, row in enumerate(data): |
| yield idx, row |
|
|
