I'm writing a document QA program that ingests a pdf document, runs an OpenAI ADA embedding over the documents and then stores the document in a vector db for later querying.
The problem is that when I run queries on the index, it gets some of the questions right, and fails at others.
An example: 'Who is the author of this book?' returns random authors from the 'references' section of the paper, when there is a sufficiently clear 'About the author section'.
What is the best practise for making your queries and results more contextually aligned?
Thank you!
