alamios/OpenSubtitles-v2024

OpenSubtitles v2024

This is a new collection of translated movie and TV subtitles from opensubtitles.org, released by the OPUS project.

IMPORTANT: If you use the OpenSubtitle corpus: Please, add a link to http://www.opensubtitles.org/ to your website and to your reports and publications produced with the data!

This is a slightly cleaner version of the subtitle collection using improved sentence alignment and better language checking.

Dataset Details

• Release: 2024 (OPUS stats: 93 languages, 3,377 bitexts, 41.03G tokens, 5.33G sentence fragments)
• This repository: all 2,997 available moses pairs, in both directions (5,994 configs), 35.2B rows (17.6B unique sentence pairs), ~1.94 TB parquet

Structure

Each config (e.g. en-es) contains a single train split with aligned sentence pairs:

{
    "id": "0",
    "translation": {
        "en": "English sentence",
        "es": "Spanish sentence"
    }
}

The format is the same as alamios/OpenSubtitles-v2018, except for the meta field (year, imdbId, subtitleId, sentenceIds), which is not available in the OPUS v2024 release.

Each pair is available in both directions for convenience:

• en-es: English as first language, Spanish as second
• es-en: Spanish as first language, English as second

Usage

from datasets import load_dataset

# Load a specific pair
ds = load_dataset("alamios/OpenSubtitles-v2024", "en-es", split="train")

# Or in streaming mode
ds = load_dataset("alamios/OpenSubtitles-v2024", "en-es", split="train", streaming=True)

for example in ds.take(5):
    print(example["translation"]["en"])
    print(example["translation"]["es"])

Available Pairs

All 2997 language pairs of the OPUS v2024 release, in both directions (5994 configs). Each config XX-YY contains the XX→YY sentences; YY-XX is the reverse direction. The 93 languages are:

af, am, an, ar, as, ast, az, be, bg, bn, br, bs, ca, cs, cy, da, de, el, en, en_ze, eo, es, es_419, et, eu, fa, fi, fr, ga, gd, gl, he, hi, hr, hu, hy, id, ig, is, it, ja, ka, kk, km, kn, ko, ku, lb, lt, lv, mk, ml, mn, mr, ms, my, ne, nl, nn, no, oc, or, pl, pt, ro, ru, sd, se, si, sk, sl, so, sq, sr, sv, sw, ta, te, th, tl, tr, tt, uk, ur, vi, yue, zh

Differences from v2018

Feature	v2018	v2024
Languages	62	93
Tokens	22.10G	41.03G
Sentence fragments	3.35G	5.33G
Metadata (year, imdbId)	Yes	No
Format	Moses (3 files)	Moses (2 files)

Citation

If you use this dataset, please cite:

P. Lison and J. Tiedemann, 2016, OpenSubtitles2016: Extracting Large Parallel Corpora from Movie and TV Subtitles. In Proceedings of the 10th International Conference on Language Resources and Evaluation (LREC 2016)

And the OPUS project:

J. Tiedemann, 2012, Parallel Data, Tools and Interfaces in OPUS. In Proceedings of the 8th International Conference on Language Resources and Evaluation (LREC 2012)

License

Unknown. The OPUS project states: "We do not own any of the text from which the data has been extracted. We only offer files that we believe we are free to redistribute."