Add the full document retrieval indexes for ar, bn, en, fi, id, ko, ru, sw, te
as sharded parquet files (2.0.0/document_retrieval_cross_lang_<lang>_index-*.parquet,
columns title/text; 373 shards, 12,321,370 articles). They are converted from
TFDS wikipedia/20190301.<lang> (version 1.0.0), which MSEB previously loaded
via TFDS.
Also add the merged TFDS dataset_info.json and the CC BY-SA 3.0 license of the
indexes, and README.md dataset viewer configs and documentation.
- Add parquet versions of the index files. (62639707d3dd2e5c1e551438539ff86c1968db16)
Co-authored-by: Georg Heigold <heigold@users.noreply.huggingface.co>
- Add version 2.0.0, which consistently uses per-locale and per-environment parquet files. (26fd6b0cff0d1ae168bfbaba5bf1906511a56cdb)
- Add version 2.0.0, which consistently uses per-locale and per-environment parquet files. (3cbd1369fe27680e5a7c18b2befed54975a51db6)
- Add version 2.0.0, which consistently uses per-locale and per-environment parquet files. (80305052e0b8bcd78843ea8fd7ba8cd4d2d4348f)
- Remove audio/ prefix from utt_index.jsonl (0c22e375167bc988be8c50461be3f443f61ae7e7)
- Update salient terms and task stats. (101a4eda4488f5c69826a4a0d4651a190478b15e)
- Add row group size and a page index. (824e5864de214fc438c3b0e6c1b9d1687f639e98)
Co-authored-by: Georg Heigold <heigold@users.noreply.huggingface.co>
- Add annotations for salient term segmentation. (124c41c6a8efa13ca369aac2977d598e0cfc08e1)
Co-authored-by: Georg Heigold <heigold@users.noreply.huggingface.co>
