itacola.py

import csv
import sys

import datasets
from typing import List

csv.field_size_limit(sys.maxsize)


_CITATION = """\
@inproceedings{trotta-etal-2021-monolingual,
    author = {Trotta, Daniela and Guarasci, Raffaele and Leonardelli, Elisa and Tonelli, Sara},
    title = {Monolingual and Cross-Lingual Acceptability Judgments with the Italian {CoLA} corpus},
    booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2021",
    month = nov,
    year = {2021},
    address = "Punta Cana, Dominican Republic and Online",
    publisher = "Association for Computational Linguistics",
    url = "https://arxiv.org/abs/2109.12053",
}
"""

_DESCRIPTION = """\
The Italian Corpus of Linguistic Acceptability includes almost 10k sentences taken from 
linguistic literature with a binary annotation made by the original authors themselves. 
The work is inspired by the English Corpus of Linguistic Acceptability (CoLA) by Warstadt et al.
Part of the dataset has been manually annotated to highlight 9 linguistic phenomena.
"""

_HOMEPAGE = "https://github.com/dhfbk/ItaCoLA-dataset"

_LICENSE = "None"

_SPLITS = ["train", "test"]


class ItaColaConfig(datasets.BuilderConfig):
    """BuilderConfig for ItaCoLA."""

    def __init__(
        self,
        features,
        data_url,
        **kwargs,
    ):
        """
        Args:
        features: `list[string]`, list of the features that will appear in the
            feature dict. Should not include "label".
        data_url: `string`, url to download the zip file from.
        **kwargs: keyword arguments forwarded to super.
        """
        super().__init__(version=datasets.Version("1.0.0"), **kwargs)
        self.data_url = data_url
        self.features = features


class ItaCola(datasets.GeneratorBasedBuilder):
    VERSION = datasets.Version("1.0.0")

    BUILDER_CONFIGS = [
        ItaColaConfig(
            name="scores",
            features=["unique_id", "source", "acceptability", "sentence"],
            data_url="https://raw.githubusercontent.com/dhfbk/ItaCoLA-dataset/main/ItaCoLA_dataset.tsv"
        ),
        ItaColaConfig(
            name="phenomena",
            features=[
                "unique_id",
                "source",
                "acceptability",
                "sentence",
                "cleft_construction",
                "copular_construction",
                "subject_verb_agreement",
                "wh_islands_violations",
                "simple",
                "question",
                "auxiliary",
                "bind",
                "indefinite_pronouns",
            ],
            data_url="https://github.com/dhfbk/ItaCoLA-dataset/raw/main/ItaCoLA_dataset_phenomenon.tsv"
        ),
    ]

    DEFAULT_CONFIG_NAME = "scores"

    def _info(self):
        features = {feature: datasets.Value("int32") for feature in self.config.features}
        features["source"] = datasets.Value("string")
        features["sentence"] = datasets.Value("string")
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=datasets.Features(features),
            homepage=_HOMEPAGE,
            license=_LICENSE,
            citation=_CITATION,
        )

    def _split_generators(self, dl_manager):
        """Returns SplitGenerators."""
        data_file = dl_manager.download_and_extract(self.config.data_url)
        if self.config.name == "scores":
            return [
                datasets.SplitGenerator(
                    name=datasets.Split.TRAIN,
                    gen_kwargs={
                        "filepath": data_file,
                        "split": "train",
                        "features": self.config.features,
                    },
                ),
                datasets.SplitGenerator(
                    name=datasets.Split.TEST,
                    gen_kwargs={
                        "filepath": data_file,
                        "split": "test",
                        "features": self.config.features,
                    },
                ),
            ]
        else:
            return [
                datasets.SplitGenerator(
                    name=datasets.Split.TRAIN,
                    gen_kwargs={
                        "filepath": data_file,
                        "split": "train",
                        "features": self.config.features,
                    },
                ),
            ]
    
    def _generate_examples(self, filepath: str, split: str, features: List[str]):
        """Yields examples as (key, example) tuples."""
        with open(filepath, encoding="utf8") as f:
            for id_, row in enumerate(f):
                if id_ == 0:
                    continue
                ex_split = None
                fields = row.strip().split("\t")
                if len(fields) < 6:
                    ex_split = fields[-1]
                    fields = fields[:-1]
                if ex_split is None or ex_split.strip() == split:
                    yield id_, {
                        k:v.strip() for k,v in zip(features, fields)
                    }