Clearly identifying a context for a Q&A model is essential to receiving accurate output. I was wondering if there are any general strategies for building a context that have proven to perform better. In my case, the data I want to query is ‘somewhat’ structured - like in this example:
Product Q&A
Price: $100
Date Listed: 12/30/2022
Description:
If I want my users to be able to ask questions about all of the data (including the unstructured description) I was seeking advice on how to best build my context. I am testing out a variety of different Q&A models so I am not limited on this front. Would fine-tuning help more than context generation?
Thanks in advance for the advice!