Hello community,
Can we perform in knime embedding with lang chain then pass the results to berttopic to create clustered topics then pass the representative documsnts to llm to determine what customers are saying. this is needed because we are dealing with both arabic and english and the knowm models performed acceptably in english but failed in arabic and dialect through several tests. if it can, your support would much appreciated
Hi mh,
I do not know langchain but are you aware that Mistral offers LLMs optimized for arabic languages?
Mistral Saba | Mistral AI
Maybe it’s worth a try.
1 Like
Hey thanks for replying i reallt appreciate it,
the issue is not with llms were using azure and its also optimized for arabic. we need a layer before llms to reduce the cost. we have more than million records that need first to be clustered then passed to LLM. I am not sure if am clear.
Hi,
Oh, unfortunately that exceeds my knowledge. But have you checked whether you might be able to use LangChain with Python? Python is very well integrated into KNIME after all.