I wanted to use the All-MiniLM-L12-v2 model in the new functionality in the Oracle 23ai database “Vector Search”.
I used it to search for data in a table containing text data in Polish. The result is good. The database found the records correctly.
Is the All-MiniLM-L12-v2 model only for use with English? Can it be used in other languages.
I ask because I may not have tested it 100% in my environment.
It looks like it’s for English…
Even if it says “English” on the model card, there are cases where it can be used for languages other than English, so please be aware of this.
It seems like it’s possible to fine-tune the model, but there may already be a similar model that supports the language you want.
In that case, it’s easier to search for it on the leaderboard, etc.
https://www.sbert.net/docs/sentence_transformer/training_overview.html
You are indeed right. I replaced the data with English (BBC News) and got much better results.
Thanks!