MultiLegalNeg.py

import json

import datasets
import pandas as pd
from huggingface_hub.file_download import hf_hub_url

try:
    import lzma as xz
except ImportError:
    import pylzma as xz

datasets.logging.set_verbosity_info()
logger = datasets.logging.get_logger(__name__)

_DESCRIPTION ="""\

    """

_HOMEPAGE =  ""

_LICENSE = ""

_CITATION = ""

_URL = {
    'data/'
}
_LANGUAGES = [
    "german", "french", "italian", "swiss", "english"
]

_ENGLISH = [
    "sherlock", "bioscope", "sfu"
]

_SHERLOCKS = [
    "dev", "test_cardboard_GOLD", "test_circle_GOLD", "training"
]

_BIOSCOPES = [
    "abstracts", "full_papers"
]


class MultiLegalNegConfig(datasets.BuilderConfig):
    
    def __init__(self, name:str, **kwargs):
        super( MultiLegalNegConfig, self).__init__(**kwargs)
        self.name = name
        self.language = name.split("_")[0]

class MultiLegalNeg(datasets.GeneratorBasedBuilder):

    BUILDER_CONFIG_CLASS = MultiLegalNegConfig
    
    BUILDER_CONFIGS = [
        MultiLegalNegConfig(f"{language}")
        for language in _LANGUAGES + ['all']
    ]
    DEFAULT_CONFIG_NAME = 'all_all'
    def _info(self):
        features = datasets.Features(
            {   
                "text": datasets.Value("string"),        
                "spans": [
                    {
                        "start": datasets.Value("int64"),
                        "end": datasets.Value("int64"),
                        "token_start": datasets.Value("int64"),
                        "token_end": datasets.Value("int64"),
                        "label": datasets.Value("string")                     
                    }
                ],
                "tokens": [
                    {
                        "text": datasets.Value("string"),
                        "start": datasets.Value("int64"),
                        "end": datasets.Value("int64"),
                        "id": datasets.Value("int64"),
                        "ws": datasets.Value("bool")
                    }
                ]
            }
        )
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features = features,
            homepage = _HOMEPAGE,
            citation=_CITATION
        )
    
    def _split_generators(self, dl_manager):
        data_files = {
            "train": [
                "data/train/it_train.jsonl.xz",
                "data/train/fr_train.jsonl.xz",
                "data/train/de_train.jsonl.xz",
                "data/train/swiss_train.jsonl.xz",
                "data/train/en_sherlock_train.jsonl.xz",
                "data/train/en_sfu_train.jsonl.xz",
                "data/train/en_bioscope_train.jsonl.xz"
            ],
            "test": [
                "data/test/it_test.jsonl.xz",
                "data/test/fr_test.jsonl.xz",
                "data/test/de_test.jsonl.xz",
                "data/test/swiss_test.jsonl.xz",
                "data/test/en_sherlock_test.jsonl.xz",
                "data/test/en_sfu_test.jsonl.xz",
                "data/test/en_bioscope_test.jsonl.xz"
            ],
            "validation": [
                "data/validation/it_validation.jsonl.xz",
                "data/validation/fr_validation.jsonl.xz",
                "data/validation/de_validation.jsonl.xz",
                "data/validation/swiss_validation.jsonl.xz",
                "data/validation/en_sherlock_validation.jsonl.xz",
                "data/validation/en_sfu_validation.jsonl.xz",
                "data/validation/en_bioscope_validation.jsonl.xz"
            ]
        }
        
        train_data = [{"text": line.strip(), "language": lang} for lang, files in data_files.items() for file in files for line in xz.open(file, "rt", encoding="utf-8")]
        test_data = [{"text": line.strip(), "language": lang} for lang, files in data_files.items() for file in files for line in xz.open(file, "rt", encoding="utf-8")]
        validation_data = [{"text": line.strip(), "language": lang} for lang, files in data_files.items() for file in files for line in xz.open(file, "rt", encoding="utf-8")]
        
        return [
            self._split_generate("train", data=train_data),
            self._split_generate("test", data=test_data),
            self._split_generate("validation", data=validation_data)
        ]

    def _split_generate(self, split, data):
        return self.DatasetSplitGenerator(
            name=split,
            gen_kwargs={"data": data},
        )

    def _generate_examples(self, data):
        for i, example in enumerate(data):
            yield i, example