vsolscsum.py

import xml.etree.ElementTree as ET
from pathlib import Path
from typing import Dict, List, Tuple

import datasets
import pandas as pd

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

_CITATION = """\
@inproceedings{nguyen-etal-2016-vsolscsum,
    title = "{VS}o{LSCS}um: Building a {V}ietnamese Sentence-Comment Dataset for Social Context Summarization",
    author = "Nguyen, Minh-Tien  and
      Lai, Dac Viet  and
      Do, Phong-Khac  and
      Tran, Duc-Vu  and
      Nguyen, Minh-Le",
    editor = "Hasida, Koiti  and
      Wong, Kam-Fai  and
      Calzorari, Nicoletta  and
      Choi, Key-Sun",
    booktitle = "Proceedings of the 12th Workshop on {A}sian Language Resources ({ALR}12)",
    month = dec,
    year = "2016",
    address = "Osaka, Japan",
    publisher = "The COLING 2016 Organizing Committee",
    url = "https://aclanthology.org/W16-5405",
    pages = "38--48",
    }
"""

_DATASETNAME = "vsolscsum"

_DESCRIPTION = """
The Vietnamese dataset for social context summarization \
    The dataset contains 141 open-domain articles along with \
    3,760 sentences, 2,448 extracted standard sentences and \
    comments as standard summaries and 6,926 comments in 12 \
    events. This dataset was manually annotated by human. \
    Note that the extracted standard summaries also include comments.\
    The label of a sentence or comment was generated based on the \
    voting among social annotators. For example, given a sentence, \
    each annotator makes a binary decision in order to indicate \
    that whether this sentence is a summary candidate (YES) or not \
    (NO). If three annotators agree yes, this sentences is labeled by 3. \
    Therefore, the label of each sentence or comment ranges from 1 to 5\
    (1: very poor, 2: poor, 3: fair, 4: good; 5: perfect). The standard \
    summary sentences are those which receive at least three agreements \
    from annotators. The inter-agreement calculated by Cohen's Kappa \
    after validation among annotators is 0.685.
"""

_HOMEPAGE = "https://github.com/nguyenlab/VSoLSCSum-Dataset"

_LANGUAGES = ["vie"]

_LICENSE = Licenses.CC_BY_4_0.value

_LOCAL = False

_URLS = {
    _DATASETNAME: "https://raw.githubusercontent.com/nguyenlab/VSoLSCSum-Dataset/master/VSoSLCSum.xml",
}

_SUPPORTED_TASKS = [Tasks.SUMMARIZATION]

_SOURCE_VERSION = "1.0.0"

_SEACROWD_VERSION = "2024.06.20"


class VSolSCSumDataset(datasets.GeneratorBasedBuilder):
    """
    The Vietnamese dataset for social context summarization includes 141 articles
    with a total of 3,760 sentences. It also contains 2,448 standard sentences
    extracted along with comments serving as standard summaries, and 6,926 c
    omments across 12 events. Human annotators manually curated this dataset.
    Each sentence or comment received a label from 1 to 5 based on annotators'
    agreement (1: very poor, 2: poor, 3: fair, 4: good, 5: perfect). Standard
    summary sentences are those with at least three agreements. The inter-agreement
    among annotators, measured by Cohen's Kappa, is 0.685.
    """

    SOURCE_VERSION = datasets.Version(_SOURCE_VERSION)
    SEACROWD_VERSION = datasets.Version(_SEACROWD_VERSION)

    BUILDER_CONFIGS = [
        SEACrowdConfig(
            name=f"{_DATASETNAME}_source",
            version=SOURCE_VERSION,
            description=f"{_DATASETNAME} source schema",
            schema="source",
            subset_id=f"{_DATASETNAME}",
        ),
        SEACrowdConfig(
            name=f"{_DATASETNAME}_seacrowd_t2t",
            version=SEACROWD_VERSION,
            description=f"{_DATASETNAME} SEACrowd schema",
            schema="seacrowd_t2t",
            subset_id=f"{_DATASETNAME}",
        ),
    ]

    DEFAULT_CONFIG_NAME = f"{_DATASETNAME}_source"

    def _info(self) -> datasets.DatasetInfo:

        if self.config.schema == "source":
            features = datasets.Features(
                {
                    "post_id": datasets.Value("string"),
                    "title": datasets.Value("string"),
                    "summary": datasets.Value("string"),
                    "document_and_comment": datasets.Value("string"),
                }
            )

        elif self.config.schema == "seacrowd_t2t":
            features = schemas.text2text_features

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

    def _split_generators(self, dl_manager: datasets.DownloadManager) -> List[datasets.SplitGenerator]:
        """Returns SplitGenerators."""

        data_path = Path(dl_manager.download_and_extract(_URLS[_DATASETNAME]))

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

    def _generate_examples(self, filepath: Path, split: str) -> Tuple[int, Dict]:
        """Yields examples as (key, example) tuples."""

        with open(filepath, "r", encoding="utf-8") as file:
            xml_content = file.read()

        root = ET.fromstring(xml_content)

        def extract_data_from_xml(root):
            data = []

            for post in root.findall(".//post"):
                post_id = post.get("id")
                title = post.find("title").text
                summary_sentences = [sentence.find("content").text for sentence in post.find(".//summary").find("sentences").findall("sentence")]
                document_sentences = [sentence.find("content").text for sentence in post.find(".//document").find("sentences").findall("sentence")]
                comment_sentences = [sentence.find("content").text for sentence in post.find(".//comments").find(".//comment").find("sentences").findall("sentence")]

                summary_text = " ".join(summary_sentences)
                document_text = " ".join(document_sentences)
                comment_text = " ".join(comment_sentences)

                data.append(
                    {
                        "post_id": post_id,
                        "title": title,
                        "summary": summary_text,
                        "document_and_comment": f"{document_text} | {comment_text}",
                    }
                )

            return data

        extracted_data = extract_data_from_xml(root)
        df = pd.DataFrame(extracted_data)

        for index, row in df.iterrows():

            if self.config.schema == "source":
                example = row.to_dict()

            elif self.config.schema == "seacrowd_t2t":

                example = {
                    "id": str(row["post_id"]),
                    "text_1": str(row["summary"]),
                    "text_2": str(row["document_and_comment"]),
                    "text_1_name": "summary",
                    "text_2_name": "document_and_comment",
                }

            yield index, example