Lets’s say you upload a scanned pdf (or not) which has tables (like a row of items) and sometimes using the chat isn’t the best. Would it be better to get the text of the PDF and insert the table row items as metadata?
Can you share more about what your queries/chat prompts look like when you say “chat isn’t the best”? That will help in figuring out the best approach or if there’s something we need to do on our end with how table data is processed.
Also, if you’d like to join the Pinecone community on Discord, there are others there that can help answer your questions in a little bit more real time. Invite: Pinecone Nearest Neighbors: Official Community
@jenna When I upload a multi-modal pdf that has scanned PDF of table columns of rows of data.. sometimes it hallucinates the answers or gets wrong answer on the Assistant. So.. I’m at an index.. whats the best way to store as a record? The whole row? The whole table? etc
For examples:
Some tables header cols and others do not.
I use https://www.docling.ai/ to clean my data before ingestion