Embedding Models
#情報検索
Benchmarks
⽇本語埋め込みモデル評価ベンチマークの構築
https://jedworkshop.github.io/JLR2024/materials/b-3.pdf
MIRACL: Multilingual Information Retrieval Across a Continuum of Languages
https://github.com/project-miracl/miracl
https://direct.mit.edu/tacl/article/doi/10.1162/tacl_a_00595/117438/MIRACL-A-Multilingual-Retrieval-Dataset-Covering
MTEB benchmark
レポ
https://github.com/embeddings-benchmark/mteb
論文
https://arxiv.org/abs/2210.07316
メモ
RetrievalとclusteringはEnglish-only (付録B.3)
Multilingual-E5
概要
論文
https://arxiv.org/pdf/2402.05672
モデル
https://huggingface.co/intfloat/multilingual-e5-base
https://huggingface.co/intfloat/multilingual-e5-large
解説
https://hironsan.hatenablog.com/entry/2023/07/05/073150
https://note.com/retrieva/n/n5abba0f8c7c4
注意点
IRなどクエリと文書が非対称な用途で使用する場合は、接頭辞にquery:とpassage:をつける
Cross-language information retrieval
https://community.openai.com/t/openai-embeddings-multi-languagm,e/557710/20
https://blog.vespa.ai/simplify-search-with-multilingual-embeddings/