Indexing embeddings when storing them in the vector database

I create embeddings with my dataset. I store the embeddings in the pinecone database and index them when upsert to the database.
The columns in my dataset are as follows [‘title’, ‘subtitle’, ‘content’, ‘faq’]. I use the title and subtitle columns when indexing embeddings, I think indexes have an effect on making more accurate mappings in semantic search. Do you think this idea is correct, or will they work with the same performance if I give them a unique random ID?

Hi @klcogluberk

To clarify, is this a question about the text you’re embedding, or a question that’s specific to Pinecone?

1 Like

When I embed a text, I need to index it to store it. In other, if I chose a different database or local methods other than pinecone, I would have to indexing embeddings likeways. So my question can be said in general. I preferably use the pinecone database.