PLOD-filtered.py

import os

import datasets
from typing import List
import json

logger = datasets.logging.get_logger(__name__)


_CITATION = """
"""

_DESCRIPTION = """
This is the dataset repository for PLOD Dataset accepted to be published at LREC 2022.
The dataset can help build sequence labelling models for the task Abbreviation Detection.
"""

class PLODfilteredConfig(datasets.BuilderConfig):
    """BuilderConfig for Conll2003"""

    def __init__(self, **kwargs):
        """BuilderConfig forConll2003.
        Args:
          **kwargs: keyword arguments forwarded to super.
        """
        super(PLODfilteredConfig, self).__init__(**kwargs)


class PLODfilteredConfig(datasets.GeneratorBasedBuilder):
    """PLOD Filtered dataset."""

    BUILDER_CONFIGS = [
        PLODfilteredConfig(name="PLODfiltered", version=datasets.Version("0.0.2"), description="PLOD filtered dataset"),
    ]

    def _info(self):
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=datasets.Features(
                {
                    "id": datasets.Value("string"),
                    "tokens": datasets.Sequence(datasets.Value("string")),
                    "pos_tags": datasets.Sequence(
                        datasets.features.ClassLabel(
                            names=[
                                "ADJ",
                                "ADP",
                                "ADV",
                                "AUX",
                                "CONJ",
                                "CCONJ",
                                "DET",
                                "INTJ",
                                "NOUN",
                                "NUM",
                                "PART",
                                "PRON",
                                "PROPN",
                                "PUNCT",
                                "SCONJ",
                                "SYM",
                                "VERB",
                                "X",
                                "SPACE"
                            ]
                        )
                    ),
                    "ner_tags": datasets.Sequence(
                        datasets.features.ClassLabel(
                            names=[
                                "B-O",
                                "B-AC",
                                "I-AC",
                                "B-LF",
                                "I-LF"
                            ]
                        )
                    ),
                }
            ),
            supervised_keys=None,
            homepage="https://github.com/surrey-nlp/PLOD-AbbreviationDetection",
            citation=_CITATION,
        )

    _URL = "https://huggingface.co/datasets/surrey-nlp/PLOD-filtered/resolve/main/data/"
    _URLS = {
        "train": _URL + "PLOS-train70-filtered-pos_bio.json",
        "dev": _URL + "PLOS-val15-filtered-pos_bio.json",
        "test": _URL + "PLOS-test15-filtered-pos_bio.json"
    }

    def _split_generators(self, dl_manager: datasets.DownloadManager) -> List[datasets.SplitGenerator]:
        urls_to_download = self._URLS
        downloaded_files = dl_manager.download_and_extract(urls_to_download)

        return [
            datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"filepath": downloaded_files["train"]}),
            datasets.SplitGenerator(name=datasets.Split.VALIDATION, gen_kwargs={"filepath": downloaded_files["dev"]}),
            datasets.SplitGenerator(name=datasets.Split.TEST, gen_kwargs={"filepath": downloaded_files["test"]})
        ]

    def _generate_examples(self, filepath):
        """This function returns the examples in the raw (text) form."""
        logger.info("generating examples from = %s", filepath)
        with open(filepath) as f:
            plod = json.load(f)
            for object in plod:
                id_ = int(object['id'])
                yield id_, {
                    "id": str(id_),
                    "tokens": object['tokens'],
                    "pos_tags": object['pos_tags'],
                    "ner_tags": object['ner_tags'],
                }