I'm currently working on a project where I need to vectorize a list of JSON objects that represent products. Unfortunately, the AI Search I'm using only allows vectorization of a single field. To tackle this, I've been concatenating all the fields into one string formatted like this: : . : . While this method has worked reasonably well for many products, I've encountered issues, particularly with queries related to product IDs. I'm seeking advice on two main points: 1) Is there an automated solution that could enhance the search quality? 2) If not, is my concatenation approach effective, or could it lead to problems down the line?
3 Answers
You might want to consult CoPilot for more specific insights regarding your implementation.
When concatenating fields, it's crucial to structure them like a coherent sentence. Depending on how your API functions, perhaps consider using multiple indices—one for each property—which might allow for more targeted searches across all fields.
That's an interesting point! The thing is, I'm vectorizing product IDs because they are critical for some queries, although it sounds odd.
It sounds like you need to treat each product as its own individual document. This way, you can directly vectorize the fields without losing any detail. It could improve your search accuracy significantly!
That makes sense, but the main challenge remains how to effectively vectorize multiple properties. Azure AI Search limits you to one vector field, right?
Actually, the current setup I have is pretty much based on inputs from CoPilot and ChatGPT already.