dataset.py

import datasets

_DESCRIPTION = "Cleaned Hong Kong Cantonese / Traditional Chinese corpora from multiple sources."
_HOMEPAGE = "https://huggingface.co/datasets/SolarisCipher/hk_content_corpus"
_CORPUS_FILES = {
    "appledaily_article": "appledaily_article_dedup.txt",
    "hkcitizenmedia_article": "hkcitizenmedia_article_dedup.txt",
    "hkcnews_article": "hkcnews_article_dedup.txt",
    "inmedia_article": "inmedia_article_dedup.txt",
    "lihkg_posts": "lihkg_posts_dedup_demoji_128.txt",
    "openrice_review": "openrice_review_dedup_demoji.txt",
    "thestandnews_article": "thestandnews_article_dedup.txt",
    "wiki_hk": "wiki_hk.txt",
}

class HKContentCorpus(datasets.GeneratorBasedBuilder):

    def _info(self):
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=datasets.Features({"text": datasets.Value("string")}),
            supervised_keys=None,
            homepage=_HOMEPAGE,
            citation="",
        )

    def _split_generators(self, dl_manager):
        return [
            datasets.SplitGenerator(
                name=split_name,
                gen_kwargs={"filepath": filepath},
            )
            for split_name, filepath in _CORPUS_FILES.items()
        ]

    def _generate_examples(self, filepath):
        with open(filepath, encoding="utf-8") as f:
            buffer = []
            example_id = 0
            for line in f:
                line = line.strip()
                if line:
                    buffer.append(line)
                else:
                    if buffer:
                        yield example_id, {"text": "\n".join(buffer)}
                        example_id += 1
                        buffer = []
            if buffer:
                yield example_id, {"text": "\n".join(buffer)}