id_stance.py

import json
from pathlib import Path
from typing import List

import datasets
import pandas as pd

from seacrowd.utils import schemas
from seacrowd.utils.configs import SEACrowdConfig
from seacrowd.utils.constants import Tasks

_CITATION = """\
@INPROCEEDINGS{8629144,  
  author={R. {Jannati} and R. {Mahendra} and C. W. {Wardhana} and M. {Adriani}},
  booktitle={2018 International Conference on Asian Language Processing (IALP)},
  title={Stance Classification Towards Political Figures on Blog Writing},
  year={2018},
  volume={},
  number={},
  pages={96-101},
}
"""

_LANGUAGES = ["ind"]  # We follow ISO639-3 language code (https://iso639-3.sil.org/code_tables/639/data)
_LOCAL = False

_DATASETNAME = "id_stance"
_DESCRIPTION = """\
Stance Classification Towards Political Figures on Blog Writing.
This dataset contains dataset from the second research, which is combined from the first research and new dataset.
The dataset consist of 337 data, about five target and every target have 1 different event.
Two label are used: 'For' and 'Againts'.
1. For - the text that is created by author is support the target in an event
2. Against - the text that is created by author is oppose the target in an event
"""
_HOMEPAGE = "https://github.com/reneje/id_stance_dataset_article-Stance-Classification-Towards-Political-Figures-on-Blog-Writing"
_LICENSE = "Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License"
_URLs = {
    _DATASETNAME: "https://raw.githubusercontent.com/reneje/id_stance_dataset_article-Stance-Classification-Towards-Political-Figures-on-Blog-Writing/master/dataset_stance_2_label_2018_building_by_rini.csv"
}
_SUPPORTED_TASKS = [Tasks.TEXTUAL_ENTAILMENT]
_SOURCE_VERSION = "1.0.0"
_SEACROWD_VERSION = "2024.06.20"


def parse_list(content):
    if (not content):
        return []
    try:
        return json.loads(content)
    except:
        return json.loads("[\"" + content[1:-1].replace("\"", "\\\"") + "\"]")


class IdStance(datasets.GeneratorBasedBuilder):
    """The ID Stance dataset is annotated with a label whether the article is in favor of the person in the context of the event"""

    BUILDER_CONFIGS = [
        SEACrowdConfig(
            name="id_stance_source",
            version=datasets.Version(_SOURCE_VERSION),
            description="IdStance source schema",
            schema="source",
            subset_id="id_stance",
        ),
        SEACrowdConfig(
            name="id_stance_seacrowd_pairs",
            version=datasets.Version(_SEACROWD_VERSION),
            description="IdStance Nusantara schema",
            schema="seacrowd_pairs",
            subset_id="id_stance",
        ),
    ]

    DEFAULT_CONFIG_NAME = "id_stance_source"

    def _info(self):
        if self.config.schema == "source":
            features = datasets.Features(
                {
                    "person": datasets.Value("string"),
                    "event": datasets.Value("string"),
                    "title": datasets.Value("string"),
                    "content": datasets.Value("string"),
                    "stance_final": datasets.Value("string"),
                }
            )
        elif self.config.schema == "seacrowd_pairs":
            features = schemas.pairs_features(["for", "against", "no"])

        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=features,
            homepage=_HOMEPAGE,
            license=_LICENSE,
            citation=_CITATION,
        )

    def _split_generators(self, dl_manager: datasets.DownloadManager) -> List[datasets.SplitGenerator]:
        data_path = Path(dl_manager.download_and_extract(_URLs[_DATASETNAME]))
        data_files = {
            "train": data_path,
        }

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

    def _generate_examples(self, filepath: Path):
        df = pd.read_csv(filepath, sep=";", header="infer", keep_default_na=False).reset_index()
        df.columns = ["id", "person", "event", "title", "content", "stance_final", ""]
        df.content = df.content.apply(parse_list)

        if self.config.schema == "source":
            for row in df.itertuples():
                ex = {
                    "person": row.person,
                    "event": row.event,
                    "title": row.title,
                    "content": " ".join(row.content),
                    "stance_final": row.stance_final
                }
                yield row.id, ex
        elif self.config.schema == "seacrowd_pairs":
            for row in df.itertuples():
                ex = {
                    "id": row.id,
                    "text_1": row.person + " | " + row.event,
                    "text_2": " ".join([row.title] + row.content),
                    "label": 'against' if row.stance_final == 'againts' else row.stance_final
                }
                yield row.id, ex
        else:
            raise ValueError(f"Invalid config: {self.config.name}")