py_legislation.py

# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0

"""

Paraguay Legislation Dataset Builder
class PY_Legislation(datasets.GeneratorBasedBuilder)

Defines the implementation of Paraguay Legislation dataset builder (GeneratorBasedBuilder).

"""
import os
import textwrap
from textwrap import TextWrapper
import datasets
import pyarrow.parquet as pq


# [>] URLS:

# URL = "https://huggingface.co/datasets/fernandoperes/py_legislation/raw/main"
_URLS = {
    "raw_text": f"./raw_text/train.parquet",
    "unlabeled_sentences": f"./unlabeled_sentences/train.parquet",
    # "labeled_sentences_train": "./labeled_sentences/train.parquet",
    # "labeled_sentences_test": "./labeled_sentences/test.parquet",
}

_obligations = {
    "cost_type": ["sin_costo", "costo_adm", "costo_directo", "otro_costo"],

    "affected_entity": ["ent_no_afectada", "empresas", "ciudadanos", "adm_publica"],

    "io_categories": [
        "prestacion_de_informacion_empresarial_y_fiscal"
        "solicitudes_de_licencias_y_otras"
        "registros_y_notificaciones"
        "solicitud_de_subsidios_y_otras"
        "disponibilidad_de_manuales_y_otras"
        "cooperacion_con_auditorías_y_otras"
        "prestacion_de_informacion_a_consumidores"
        "otras_OIS"
    ],

    "aa_categories_unique": [
        "familiarizacion_con_OI"
        "recoleccion_y_organizacion_de_informacion"
        "procesamiento_de_informacion"
        "tiempos_de_espera"
        "desplazamientos"
        "envio_de_informacion"
        "preservacion_de_informacion"
    ],

    "aa_categories": [
        "aa_1_familiarizacion_con_OI"
        "aa_1_recoleccion_y_organizacion_de_informacion"
        "aa_1_procesamiento_de_informacion"
        "aa_1_tiempos_de_espera"
        "aa_1_desplazamientos"
        "aa_1_envio_de_informacion"
        "aa_1_preservacion_de_informacion"
        "aa_2_familiarizacion_con_OI"
        "aa_2_recoleccion_y_organizacion_de_informacion"
        "aa_2_procesamiento_de_informacion"
        "aa_2_tiempos_de_espera"
        "aa_2_desplazamientos"
        "aa_2_envio_de_informacion"
        "aa_2_preservacion_de_informacion"
        "aa_3_familiarizacion_con_OI"
        "aa_3_recoleccion_y_organizacion_de_informacion"
        "aa_3_procesamiento_de_informacion"
        "aa_3_tiempos_de_espera"
        "aa_3_desplazamientos"
        "aa_3_envio_de_informacion"
        "aa_3_preservacion_de_informacion"
        "aa_4_familiarizacion_con_OI"
        "aa_4_recoleccion_y_organizacion_de_informacion"
        "aa_4_procesamiento_de_informacion"
        "aa_4_tiempos_de_espera"
        "aa_4_desplazamientos"
        "aa_4_envio_de_informacion"
        "aa_4_preservacion_de_informacion"
        "aa_5_familiarizacion_con_OI"
        "aa_5_recoleccion_y_organizacion_de_informacion"
        "aa_5_procesamiento_de_informacion"
        "aa_5_tiempos_de_espera"
        "aa_5_desplazamientos"
        "aa_5_envio_de_informacion"
        "aa_5_preservacion_de_informacion"
        "aa_6_familiarizacion_con_OI"
        "aa_6_recoleccion_y_organizacion_de_informacion"
        "aa_6_procesamiento_de_informacion"
        "aa_6_tiempos_de_espera"
        "aa_6_desplazamientos"
        "aa_6_envio_de_informacion"
        "aa_6_preservacion_de_informacion"
        "aa_7_familiarizacion_con_OI"
        "aa_7_recoleccion_y_organizacion_de_informacion"
        "aa_7_procesamiento_de_informacion"
        "aa_7_tiempos_de_espera"
        "aa_7_desplazamientos"
        "aa_7_envio_de_informacion"
        "aa_7_preservacion_de_informacion"
    ]
}

_metadata = {
    "citation": """\
    @InProceedings{
        huggingface:dataset,
        title = {Paraguay Legislation Dataset},
        author={Peres, Fernando; Costa, Victor},
        year={2023}
    }
    """,

    "description": textwrap.dedent("""\
        Dataset for researching - NLP techniques on PARAGUAY legislation.
    """),

    "homepage": "https://www.leyes.com.py/",

    "license": "apache-2.0",
}

# [@] Config Names:
_CONFIGS = {
    "raw_text": {
        "description": textwrap.dedent("""
        Data extracted from the sources files (URls, PDFs and Word files) without any transformation or sentence splitter. It can be helpful because you can access the raw data extracted from the seeds (PDFs and Word files) and apply other preprocessing tasks from this point to prepare the data without returning to extract texts from source files.
        """),

        "features": {
            "source_id": datasets.Value(dtype="int64"),
            "source_name": datasets.Value(dtype="string"),
            "text_id": datasets.Value(dtype="int64"),
            "text": datasets.Value(dtype="string"),

            "extension": datasets.ClassLabel(names=["docx", "pdf", "html", "txt", "doc"],),
        }

    },

    "unlabeled_sentences": {
        "description": textwrap.dedent("""
        Unlabeled corpora of Paraguay legislation. This data is prepared to be labeled by the experts. Each instance of the dataset represents a specific text passage, split by its original formatting extracted from raw text (from original documents)

        Each observation of the dataset represents a specific text passage.
        """),

        "features": {
            "source_id": datasets.Value(dtype="int64"),
            "source_name": datasets.Value(dtype="string"),
            "text_id": datasets.Value(dtype="int64"),
            "text": datasets.Value(dtype="string"),

            # Categories
            "cost_type": datasets.ClassLabel(
                names=_obligations["cost_type"],),
            "affected_entity": datasets.ClassLabel(
                names=_obligations["affected_entity"],),
            "io_categories": datasets.Sequence(
                datasets.ClassLabel(names=_obligations["io_categories"],)),
            "aa_categories": datasets.Sequence(
                datasets.ClassLabel(names=_obligations["aa_categories"],)),
            "aa_categories_unique": datasets.Sequence(
                datasets.ClassLabel(names=_obligations["aa_categories_unique"],)),
        }
    },

    "labeled_sentences": {
        "description": textwrap.dedent("""
        The labeled data is the ground truth data used to train the models. This data is annotated  by legal experts indicating the existence of administrative costs (and other types) in the legislation.

        Each observation of the dataset represents a specific text passage.
        """),

        "features": {
            "source_id": datasets.Value(dtype="int64"),
            "source_name": datasets.Value(dtype="string"),
            "text_id": datasets.Value(dtype="int64"),
            "text": datasets.Value(dtype="string"),

            # Categories
            "cost_type": datasets.ClassLabel(
                names=_obligations["cost_type"],),
            "affected_entity": datasets.ClassLabel(
                names=_obligations["affected_entity"],),
            "io_categories": datasets.Sequence(
                datasets.ClassLabel(names=_obligations["io_categories"],)),
            "aa_categories": datasets.Sequence(
                datasets.ClassLabel(names=_obligations["aa_categories"],)),
            "aa_categories_unique": datasets.Sequence(
                datasets.ClassLabel(names=_obligations["aa_categories_unique"],)),
        }
    }
}


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

    BUILDER_CONFIGS = [
        datasets.BuilderConfig(
            name="raw_text",
            version=VERSION,
            description=_CONFIGS["raw_text"]["description"],
        ),

        datasets.BuilderConfig(
            name="unlabeled_sentences",
            version=VERSION,
            description=_CONFIGS["unlabeled_sentences"]["description"],
        ),

        # datasets.BuilderConfig(
        #     name="labeled_sentences",
        #     version=VERSION,
        #     description=_CONFIGS["labeled_sentences"]["description"],
        # ),
    ]

    # It's not mandatory to have a default configuration. Just use one if it make sense.
    DEFAULT_CONFIG_NAME = "raw_text"

    # [i] Dataset Information:
    def _info(self):
        """
        This method specifies the datasets.DatasetInfo object which contains
        information and typings for the dataset
        """
        features = None
        description = None

        if self.config.name in _CONFIGS.keys():
            features = datasets.Features(
                _CONFIGS[self.config.name]["features"])

            description = _CONFIGS[self.config.name]["description"]
        else:
            features = datasets.Features(
                _CONFIGS["raw_text"]["features"])

            description = _CONFIGS["raw_text"]["description"]

        return datasets.DatasetInfo(
            builder_name=self.config.name,
            description=description,
            features=features,
            homepage=_metadata["homepage"],
            license=_metadata["license"],
            citation=_metadata["citation"],
        )

    def _split_generators(self, dl_manager):
        urls = _URLS[self.config.name]
        # data_dir = dl_manager.download_and_extract(urls)
        urls = dl_manager.download_and_extract(urls)

        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={"filepath": urls},
            ),
        ]

        # region old code
        # for k in urls_to_download:
        #     fp = dl_manager.download_and_extract(urls_to_download[k])
        #     print(f"***************>>>> filepath = {fp}")
        #     generators.append(
        #         datasets.SplitGenerator(
        #             name=datasets.Split.TRAIN, gen_kwargs={"filepath": fp}))

        # us_train = dl_manager.download_and_extract(
        #     urls_to_download["unlabeled_sentences"])

        # ls_train = dl_manager.download_and_extract(
        #     urls_to_download["labeled_sentences_train"])

        # ls_test = dl_manager.download_and_extract(
        #     urls_to_download["labeled_sentences_test"])

        # generators.append(
        #    datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"filepath": us_train}))

        # datasets.SplitGenerator(
        #     name=datasets.Split.TRAIN,
        #     gen_kwargs={
        #         # downloaded_files["labeled_sentences_train"],
        #         "filepath": ls_train
        #     },
        # ),

        # datasets.SplitGenerator(
        #     name=datasets.Split.TEST,
        #     gen_kwargs={
        #         # downloaded_files["labeled_sentences_test"],
        #         "filepath": ls_test
        #     },
        # )
        # ]

        # return generators
        # endregion

    def _generate_examples(self, filepath):
        """
        This method handles input defined in _split_generators to yield (key, example) tuples from the dataset.
        The `key` is for legacy reasons (tfds) and is not important in itself, but must be unique for each example.
        Obs: method parameters are unpacked from `gen_kwargs` as given in `_split_generators`.
        """
        # # Ensure that the filepath is an absolute local file path
        # filepath = os.path.abspath(filepath)

        # # Check if the file exists
        # if not os.path.exists(filepath):
        #     raise FileNotFoundError(f"File not found: {filepath}")

        pq_table = pq.read_table(filepath)
        for i in range(len(pq_table)):
            yield i, {
                col_name: pq_table[col_name][i].as_py()
                for col_name in pq_table.column_names
            }