I need help with my thesis.

[nick-dask]

I'm trying to build a document based chatbot, which is part of my final thesis, that will later be integrated in a website that I will create, basically like a widget and will respond to any questions the user asks about the given document (the doc contains basic information about my institution). I mainly work with python and have created a viable model using langchain to build the language model and streamlit to attach the document and display the chat, but I've run into a problem. The document is in Greek and most language models do not support this specific language completely, if at all. The chatbot's answers, are totally different semantically, from the questions asked and I'm stuck with an unusable bot. What can I do to fix this? Is there any way to train the bot for my specific use case?

[jorge2najo]

I found this on the internet. I hope it can help you.
https://medium.com/deepset-ai/going-beyond-squad-part-1-question-answering-in-different-languages-8eac6cf56f21