NilanE's picture
Update README.md
d0ec442 verified
|
raw
history blame
No virus
2.25 kB
metadata
license: apache-2.0
task_categories:
  - translation
language:
  - ja
  - en

Dataset details:

Each entry in this dataset is a sentence-aligned Japanese web novel chapter and English fan translation. The intended use-case is for document translation tasks.

Dataset format:

{
    'src': 'JAPANESE WEB NOVEL CHAPTER',
    'trg': 'CORRESPONDING ENGLISH TRANSLATION',
    'meta': {
        'general': {
            'series_title_eng': 'ENGLISH SERIES TITLE',
            'series_title_jap': 'JAPANESE SERIES TITLE',
            'sentence_alignment_score': 'ALIGNMENT SCORE'
        },
        'novelupdates': {
            'link': 'NOVELUPDATES URL',
            'genres': 'NOVELUPDATES GENRES',
            'tags': 'NOVELUPDATES TAGS (think sub-genres)',
            'rating': 'NOVELUPDATES RATING (X/5)',
            'rating_votes': 'NOVELUPDATES RATING VOTES'
        },
        'syosetu': {
            'link': 'SYOSETU URL',
            'series_active': 'IS THE SERIES STILL UP ON SYOSETU (is false for 3 series, each one has no syosetu metadata beyond the link and active status)',
            'writer': 'AUTHOR'S NAME ON SYOSETU',
            'fav_novel_cnt': 'FROM SYOSETU API FOR CHECKING SERIES QUALITY',
            'global_points': 'ALSO FROM SYOSETU API FOR CHECKING SERIES QUALITY'
        }

    }
}

This is version 2 of the dataset. It contains more chapters (103K -> 106K), but has slightly fewer tokens due to an overhaul of the alignment code. This version should fix the issues found in discussions #3 and #4, adds series-specific metadata as requested in #1, and does not remove chapter titles. No translation quality filtering has been applied to the dataset. Methods for doing so are being researched.

License note:

The texts and site-specific metadata is distributed under fair use principles, with everything else being under an Apache 2.0 license. If an author, translator or one of the sites mentioned above requests a takedown of one or more series, it will be promptly addressed. Takedowns can be requested through the creation of a Huggingface disscussion.

I am not a lawyer, and the above notice is probably not legally sound. As such, I recommend discretion when using the contents of the dataset.