You have successfully imported documents in your question-answering project and you want now to give a try.
- Go to the Question answering view
- Ask a question in natural language
- The result list from the retriever is displayed first. This is the list of segments sorted by relevancy (ranking score).
- The generated answer with the LLM is presented in the green box with active link to the references in []
- You can save your question by clicking on the icon
- You can easily access all your saved / imported questions on the left panel
- Click on Show options to change configuration or customize your question answering solution
- Select your search strategy to retrieve segments: you can use full-text search, vector search or hybrid search. Hybrid search combines both full-text and vector search.
- Select your vectorizer or create a new one to experiment vector search
- Note that the default vectorizer is:
- all-MiniLM-L6-v2 for english project
- CamemBERT for french project
- Multilingual-MiniLM-L12-v2 for all languages except english and french
- Multilingual-MiniLM-L12-v2 for multilingual project
- Select your answer generator with LLMs or create a new one to experiment question answering
- Note that in general the default answer generator uses OpenAI/GPT3.5 as LLM
