Hello i am currently trying to query based off of recipe documents in my database. I have vectorized each document through the Pinecone vector store, although when i query a request such as, "salmon" for recipes i get low scores for the matches although there are plenty of similar recipes in the database. I wonder how i can get higher relevance scores?
Take the scores with a grain of salt, they don't mean too much. Especially when you are comparing a single word to an entire document
Every embedding model will also have slightly different scores that are considered "good". For example, openai scores are usually arond 0.7->0.8. BGE models might be around 0.5
