mrbesher/massive-tr

MASSIVE 1.1 Turkish (tr-TR), parquet mirror

Faithful parquet mirror of the Turkish split of MASSIVE 1.1 (Amazon Science), built from the official release tarball. The upstream Hub dataset AmazonScience/massive uses a Python loading script, which datasets>=4.0 can no longer execute; this mirror restores plain load_dataset access for the Turkish subset.

• Source: amazon-massive-dataset-1.1.tar.gz (official S3 release), file 1.1/data/tr-TR.jsonl
• Schema: identical to the original loader, including intent/scenario ClassLabel id order copied verbatim from massive.py, so ids match the script-based dataset exactly
• Splits: train 11,514 / validation 2,033 / test 2,974 (original dev = validation here)
• Annotations: utt (raw utterance), annot_utt (bracket-annotated slots, e.g. [date : bu hafta] beni [time : sabah beşte] uyandır), 60 intents, 18 scenarios, per-row worker judgments
• Build script: mirror_massive_tr.py in the training repo; deterministic, no filtering, no reordering

License & attribution

CC-BY-4.0, © Amazon.com Inc. or its affiliates, unchanged from the original release. If you use this data, cite the MASSIVE paper:

@inproceedings{fitzgerald-etal-2023-massive,
    title = "{MASSIVE}: A 1{M}-Example Multilingual Natural Language Understanding Dataset with 51 Typologically-Diverse Languages",
    author = "FitzGerald, Jack and Hench, Christopher and Peris, Charith and others",
    booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics",
    year = "2023",
    url = "https://aclanthology.org/2023.acl-long.235",
}