how to use fuzzy search in weaviate. Currently is there any method which used fuzzy logic. I want to get similar results from vector db for particular question text, if data or text stored in db is somehow different from one I am actually passing then i am getting different result. I have strored data as a paragraphs in vector db any i am passing some information in the question related to that paragraph, so i think to get with this fuzzy search is helpful?
also i want to conform that weaviate returns results in decreasing order of certainity, is it correct?
With the hybrid search, you will be able to get similar results (based on vector search) as well as keyword search. It will perform those two searches and fuse it’s results.
And yes, by default, it will return objects based on similarity (score, in hybrid and bm25) and distance (for vector search)
Thanks for the question, Abhishek, and for the helpful explanation about hybrid search.
Since your data is stored as paragraphs, I’d first try splitting longer paragraphs into smaller, meaningful chunks. Then use hybrid search and adjust alpha to balance semantic similarity against keyword matching.
I’d also inspect the returned scores to see whether the wrong results come from poor retrieval or overly broad text chunks. For vector search, a lower distance means a closer match; hybrid search uses a fused score.
Levenshtein distance can help with spelling mistakes, but it won’t solve differences in meaning between whole paragraphs.
This explanation of approximate string matching may help clarify the distinction:
For transparency, I contribute to the linked reference.
