--- dataset_info: - config_name: hpprc_emb__auto-wiki-nli-triplet features: - name: anchor dtype: string - name: positive dtype: string - name: negative_1 dtype: string - name: negative_2 dtype: string - name: negative_3 dtype: string - name: negative_4 dtype: string - name: negative_5 dtype: string - name: negative_6 dtype: string - name: negative_7 dtype: string - name: negative_8 dtype: string splits: - name: train num_bytes: 297860496 num_examples: 198895 download_size: 206911712 dataset_size: 297860496 - config_name: hpprc_emb__auto-wiki-qa features: - name: anchor dtype: string - name: positive dtype: string - name: negative_1 dtype: string - name: negative_2 dtype: string - name: negative_3 dtype: string - name: negative_4 dtype: string - name: negative_5 dtype: string - name: negative_6 dtype: string - name: negative_7 dtype: string - name: negative_8 dtype: string splits: - name: train num_bytes: 10506898171 num_examples: 1313686 download_size: 5708581136 dataset_size: 10506898171 - config_name: hpprc_emb__auto-wiki-qa-nemotron features: - name: anchor dtype: string - name: positive dtype: string - name: negative_1 dtype: string - name: negative_2 dtype: string - name: negative_3 dtype: string - name: negative_4 dtype: string - name: negative_5 dtype: string - name: negative_6 dtype: string - name: negative_7 dtype: string - name: negative_8 dtype: string splits: - name: train num_bytes: 2823115518 num_examples: 293596 download_size: 1628014643 dataset_size: 2823115518 - config_name: hpprc_emb__auto-wiki-qa-pair features: - name: anchor dtype: string - name: positive dtype: string splits: - name: train num_bytes: 1428864676 num_examples: 2377503 download_size: 799686336 dataset_size: 1428864676 - config_name: hpprc_emb__baobab-wiki-retrieval features: - name: anchor dtype: string - name: positive dtype: string - name: negative_1 dtype: string - name: negative_2 dtype: string - name: negative_3 dtype: string - name: negative_4 dtype: string - name: negative_5 dtype: string - name: negative_6 dtype: string - name: negative_7 dtype: string - name: negative_8 dtype: string splits: - name: train num_bytes: 13509043 num_examples: 1451 download_size: 8133524 dataset_size: 13509043 - config_name: hpprc_emb__jagovfaqs features: - name: anchor dtype: string - name: positive dtype: string - name: negative_1 dtype: string - name: negative_2 dtype: string - name: negative_3 dtype: string - name: negative_4 dtype: string - name: negative_5 dtype: string - name: negative_6 dtype: string - name: negative_7 dtype: string - name: negative_8 dtype: string splits: - name: train num_bytes: 123137056 num_examples: 17686 download_size: 55230398 dataset_size: 123137056 - config_name: hpprc_emb__janli-triplet features: - name: anchor dtype: string - name: positive dtype: string - name: negative_1 dtype: string - name: negative_2 dtype: string - name: negative_3 dtype: string - name: negative_4 dtype: string - name: negative_5 dtype: string - name: negative_6 dtype: string - name: negative_7 dtype: string - name: negative_8 dtype: string splits: - name: train num_bytes: 4120024 num_examples: 6775 download_size: 1128973 dataset_size: 4120024 - config_name: hpprc_emb__jaquad features: - name: anchor dtype: string - name: positive dtype: string - name: negative_1 dtype: string - name: negative_2 dtype: string - name: negative_3 dtype: string - name: negative_4 dtype: string - name: negative_5 dtype: string - name: negative_6 dtype: string - name: negative_7 dtype: string - name: negative_8 dtype: string splits: - name: train num_bytes: 2102981137 num_examples: 170557 download_size: 1002526294 dataset_size: 2102981137 - config_name: hpprc_emb__jqara features: - name: anchor dtype: string - name: positive dtype: string - name: negative_1 dtype: string - name: negative_2 dtype: string - name: negative_3 dtype: string - name: negative_4 dtype: string - name: negative_5 dtype: string - name: negative_6 dtype: string - name: negative_7 dtype: string - name: negative_8 dtype: string splits: - name: train num_bytes: 233644618 num_examples: 25159 download_size: 122614055 dataset_size: 233644618 - config_name: hpprc_emb__jsnli-triplet features: - name: anchor dtype: string - name: positive dtype: string - name: negative_1 dtype: string - name: negative_2 dtype: string - name: negative_3 dtype: string - name: negative_4 dtype: string - name: negative_5 dtype: string - name: negative_6 dtype: string - name: negative_7 dtype: string - name: negative_8 dtype: string splits: - name: train num_bytes: 93194442 num_examples: 144190 download_size: 49222111 dataset_size: 93194442 - config_name: hpprc_emb__jsquad features: - name: anchor dtype: string - name: positive dtype: string - name: negative_1 dtype: string - name: negative_2 dtype: string - name: negative_3 dtype: string - name: negative_4 dtype: string - name: negative_5 dtype: string - name: negative_6 dtype: string - name: negative_7 dtype: string - name: negative_8 dtype: string splits: - name: train num_bytes: 3427334974 num_examples: 321909 download_size: 1653267523 dataset_size: 3427334974 - config_name: hpprc_emb__miracl features: - name: anchor dtype: string - name: positive dtype: string - name: negative_1 dtype: string - name: negative_2 dtype: string - name: negative_3 dtype: string - name: negative_4 dtype: string - name: negative_5 dtype: string - name: negative_6 dtype: string - name: negative_7 dtype: string - name: negative_8 dtype: string splits: - name: train num_bytes: 35927001 num_examples: 6417 download_size: 21926050 dataset_size: 35927001 - config_name: hpprc_emb__mkqa features: - name: anchor dtype: string - name: positive dtype: string - name: negative_1 dtype: string - name: negative_2 dtype: string - name: negative_3 dtype: string - name: negative_4 dtype: string - name: negative_5 dtype: string - name: negative_6 dtype: string - name: negative_7 dtype: string - name: negative_8 dtype: string splits: - name: train num_bytes: 175105871 num_examples: 16720 download_size: 88615319 dataset_size: 175105871 - config_name: hpprc_emb__mkqa-triplet features: - name: anchor dtype: string - name: positive dtype: string - name: negative_1 dtype: string - name: negative_2 dtype: string - name: negative_3 dtype: string - name: negative_4 dtype: string - name: negative_5 dtype: string - name: negative_6 dtype: string - name: negative_7 dtype: string - name: negative_8 dtype: string splits: - name: train num_bytes: 2845404 num_examples: 6758 download_size: 1873364 dataset_size: 2845404 - config_name: hpprc_emb__mmarco features: - name: anchor dtype: string - name: positive dtype: string - name: negative_1 dtype: string - name: negative_2 dtype: string - name: negative_3 dtype: string - name: negative_4 dtype: string - name: negative_5 dtype: string - name: negative_6 dtype: string - name: negative_7 dtype: string - name: negative_8 dtype: string splits: - name: train num_bytes: 1161466356 num_examples: 282692 download_size: 634484254 dataset_size: 1161466356 - config_name: hpprc_emb__mr-tydi features: - name: anchor dtype: string - name: positive dtype: string - name: negative_1 dtype: string - name: negative_2 dtype: string - name: negative_3 dtype: string - name: negative_4 dtype: string - name: negative_5 dtype: string - name: negative_6 dtype: string - name: negative_7 dtype: string - name: negative_8 dtype: string splits: - name: train num_bytes: 20183544 num_examples: 3467 download_size: 12878323 dataset_size: 20183544 - config_name: hpprc_emb__nu-mnli-triplet features: - name: anchor dtype: string - name: positive dtype: string - name: negative_1 dtype: string - name: negative_2 dtype: string - name: negative_3 dtype: string - name: negative_4 dtype: string - name: negative_5 dtype: string - name: negative_6 dtype: string - name: negative_7 dtype: string - name: negative_8 dtype: string splits: - name: train num_bytes: 82645507 num_examples: 77785 download_size: 52853759 dataset_size: 82645507 - config_name: hpprc_emb__nu-snli-triplet features: - name: anchor dtype: string - name: positive dtype: string - name: negative_1 dtype: string - name: negative_2 dtype: string - name: negative_3 dtype: string - name: negative_4 dtype: string - name: negative_5 dtype: string - name: negative_6 dtype: string - name: negative_7 dtype: string - name: negative_8 dtype: string splits: - name: train num_bytes: 73345261 num_examples: 109154 download_size: 38625788 dataset_size: 73345261 - config_name: hpprc_emb__paws-x-triplet features: - name: anchor dtype: string - name: positive dtype: string - name: negative_1 dtype: string - name: negative_2 dtype: string - name: negative_3 dtype: string - name: negative_4 dtype: string - name: negative_5 dtype: string - name: negative_6 dtype: string - name: negative_7 dtype: string - name: negative_8 dtype: string splits: - name: train num_bytes: 31913244 num_examples: 21684 download_size: 20717027 dataset_size: 31913244 - config_name: hpprc_emb__quiz-no-mori features: - name: anchor dtype: string - name: positive dtype: string - name: negative_1 dtype: string - name: negative_2 dtype: string - name: negative_3 dtype: string - name: negative_4 dtype: string - name: negative_5 dtype: string - name: negative_6 dtype: string - name: negative_7 dtype: string - name: negative_8 dtype: string splits: - name: train num_bytes: 1629203468 num_examples: 142762 download_size: 848475317 dataset_size: 1629203468 - config_name: hpprc_emb__quiz-works features: - name: anchor dtype: string - name: positive dtype: string - name: negative_1 dtype: string - name: negative_2 dtype: string - name: negative_3 dtype: string - name: negative_4 dtype: string - name: negative_5 dtype: string - name: negative_6 dtype: string - name: negative_7 dtype: string - name: negative_8 dtype: string splits: - name: train num_bytes: 1463156382 num_examples: 136225 download_size: 742931335 dataset_size: 1463156382 - config_name: hpprc_emb__snow-triplet features: - name: anchor dtype: string - name: positive dtype: string - name: negative_1 dtype: string - name: negative_2 dtype: string - name: negative_3 dtype: string - name: negative_4 dtype: string - name: negative_5 dtype: string - name: negative_6 dtype: string - name: negative_7 dtype: string - name: negative_8 dtype: string splits: - name: train num_bytes: 37090439 num_examples: 62758 download_size: 22040932 dataset_size: 37090439 - config_name: hpprc_llmjp-kaken features: - name: anchor dtype: string - name: positive dtype: string splits: - name: train num_bytes: 5016216312 num_examples: 1114379 download_size: 1878938416 dataset_size: 5016216312 - config_name: hpprc_llmjp_warp_html features: - name: anchor dtype: string - name: positive dtype: string splits: - name: train num_bytes: 917623809.1817185 num_examples: 209240 download_size: 519320975 dataset_size: 917623809.1817185 - config_name: hpprc_mqa_ja features: - name: anchor dtype: string - name: positive dtype: string splits: - name: train num_bytes: 3767232428 num_examples: 5826275 download_size: 1716039771 dataset_size: 3767232428 - config_name: hpprc_msmarco_ja features: - name: anchor dtype: string - name: positive dtype: string - name: negative_1 dtype: string - name: negative_2 dtype: string - name: negative_3 dtype: string - name: negative_4 dtype: string - name: negative_5 dtype: string - name: negative_6 dtype: string - name: negative_7 dtype: string splits: - name: train num_bytes: 1249778645 num_examples: 350452 download_size: 747166841 dataset_size: 1249778645 configs: - config_name: hpprc_emb__auto-wiki-nli-triplet data_files: - split: train path: hpprc_emb__auto-wiki-nli-triplet/train-* - config_name: hpprc_emb__auto-wiki-qa data_files: - split: train path: hpprc_emb__auto-wiki-qa/train-* - config_name: hpprc_emb__auto-wiki-qa-nemotron data_files: - split: train path: hpprc_emb__auto-wiki-qa-nemotron/train-* - config_name: hpprc_emb__auto-wiki-qa-pair data_files: - split: train path: hpprc_emb__auto-wiki-qa-pair/train-* - config_name: hpprc_emb__baobab-wiki-retrieval data_files: - split: train path: hpprc_emb__baobab-wiki-retrieval/train-* - config_name: hpprc_emb__jagovfaqs data_files: - split: train path: hpprc_emb__jagovfaqs/train-* - config_name: hpprc_emb__janli-triplet data_files: - split: train path: hpprc_emb__janli-triplet/train-* - config_name: hpprc_emb__jaquad data_files: - split: train path: hpprc_emb__jaquad/train-* - config_name: hpprc_emb__jqara data_files: - split: train path: hpprc_emb__jqara/train-* - config_name: hpprc_emb__jsnli-triplet data_files: - split: train path: hpprc_emb__jsnli-triplet/train-* - config_name: hpprc_emb__jsquad data_files: - split: train path: hpprc_emb__jsquad/train-* - config_name: hpprc_emb__miracl data_files: - split: train path: hpprc_emb__miracl/train-* - config_name: hpprc_emb__mkqa data_files: - split: train path: hpprc_emb__mkqa/train-* - config_name: hpprc_emb__mkqa-triplet data_files: - split: train path: hpprc_emb__mkqa-triplet/train-* - config_name: hpprc_emb__mmarco data_files: - split: train path: hpprc_emb__mmarco/train-* - config_name: hpprc_emb__mr-tydi data_files: - split: train path: hpprc_emb__mr-tydi/train-* - config_name: hpprc_emb__nu-mnli-triplet data_files: - split: train path: hpprc_emb__nu-mnli-triplet/train-* - config_name: hpprc_emb__nu-snli-triplet data_files: - split: train path: hpprc_emb__nu-snli-triplet/train-* - config_name: hpprc_emb__paws-x-triplet data_files: - split: train path: hpprc_emb__paws-x-triplet/train-* - config_name: hpprc_emb__quiz-no-mori data_files: - split: train path: hpprc_emb__quiz-no-mori/train-* - config_name: hpprc_emb__quiz-works data_files: - split: train path: hpprc_emb__quiz-works/train-* - config_name: hpprc_emb__snow-triplet data_files: - split: train path: hpprc_emb__snow-triplet/train-* - config_name: hpprc_llmjp-kaken data_files: - split: train path: hpprc_llmjp-kaken/train-* - config_name: hpprc_llmjp_warp_html data_files: - split: train path: hpprc_llmjp_warp_html/train-* - config_name: hpprc_mqa_ja data_files: - split: train path: hpprc_mqa_ja/train-* - config_name: hpprc_msmarco_ja data_files: - split: train path: hpprc_msmarco_ja/train-* license: unknown language: - ja --- - 日本語のデータセットを [SentenceTransformes](SentenceTransformer) で[学習しやすいカラム名と構造](https://sbert.net/docs/sentence_transformer/loss_overview.html)に変換したもの。 - 主に `(anchor, positive)`, `(anchor, positive, negative)`, `(anchor, positive, negative_1, ..., negative_n)` といった構造になっているため、とりわけ対照学習で使いやすくなっています。 - 以下のデータセットから作成 - https://huggingface.co/datasets/hpprc/emb - https://huggingface.co/datasets/hotchpotch/hpprc_emb-scores のリランカースコアを用いて、positive(>=0.7) / negative(<=0.3) のフィルタリングを行った - https://huggingface.co/datasets/hpprc/llmjp-kaken - https://huggingface.co/datasets/hpprc/msmarco-ja - https://huggingface.co/datasets/hotchpotch/msmarco-ja-hard-negatives のリランカースコアを用いて、positive(>=0.7) / negative(<=0.3) のフィルタリングを行った - https://huggingface.co/datasets/hpprc/mqa-ja - https://huggingface.co/datasets/hpprc/llmjp-warp-html ## 謝辞 大元のデータセットを公開している方々、日本語データセットを使いやすくまとめてくださっている [@hpprc](https://huggingface.co/hpprc) 氏、ありがとうございます。 ## ライセンス ライセンスは、各々の大元のデータセットのライセンスに従います。