LaBSE
A cross-lingual sentence encoding model proposed by Google in 2020, supporting 109 languages. It uses contrastive learning to optimize language-agnostic representations, mapping semantically equivalent content expressed in different languages to nearby positions in a shared vector space.
Timeline (last 90 days)
LaBSE (Language-agnostic BERT Sentence Embedding) was proposed by Google in 2020 and supports 109 languages.
Portuguese and Spanish are the closest pair among the four languages, with a LaBSE distance of only 0.0229.
French and Italian are the most distant pair, with a LaBSE distance of 0.0338.
Among the six language pairs, LaBSE's ranking results agreed with mBERT's ranking results on 4 of them.
现代多语言模型如mT5、BLOOM和Llama系列在数百种语言的文本上联合训练，通过共享词汇表和跨语言注意力机制构建跨语言语义地图
All Facts (5)
LaBSE (Language-agnostic BERT Sentence Embedding) was proposed by Google in 2020 and supports 109 languages.
50%UnverifiedPortuguese and Spanish are the closest pair among the four languages, with a LaBSE distance of only 0.0229.
50%UnverifiedFrench and Italian are the most distant pair, with a LaBSE distance of 0.0338.
50%UnverifiedAmong the six language pairs, LaBSE's ranking results agreed with mBERT's ranking results on 4 of them.
50%Unverified现代多语言模型如mT5、BLOOM和Llama系列在数百种语言的文本上联合训练，通过共享词汇表和跨语言注意力机制构建跨语言语义地图
50%