search

danielinux7 / Abkhaz-NLP-Data-Pipeline

Abkhazian language focused multilingual and monolingual corpuses for Natural Language Processing(NLP)
https://bagrat.space/
Creative Commons Zero v1.0 Universal
20 stars 3 forks source link

XLM-R with Abkhaz #107

Open Bachstelze opened 7 months ago

Bachstelze commented 7 months ago

Glot500 is a Roberta model that supports Abkhaz. Such a model can be used for various NLP tasks like machine translation as EncoderDecoderModel.

danielinux7 commented 5 months ago

I just tried Glot500 on a simple mask for Abkhazian, it failed to give me an answer.

It seems Claude opus 3 has been trained on the Circassian langauage and gives good results, which is a sister language for Abkhazian, I don't know if Abkhazian is trained as well, because I don't have access to Claude in my region.