TLDR Gerald experienced odd search results, and Jason suggested setting `drop_tokens_threshold` and using the curation feature to filter out certain results.
I have a circumstance where results appear to rank oddly. Let’s say I’m searching by 4 fields: primary category (furniture, etc), secondary (type), material, title. If I search for “metal chair” the result pulls up entries with the word “metal” in the title. Even though the primary and secondary category are weighted higher than title.
I suspect this is `drop_tokens_threshold` kicking in. Could you try setting `drop_tokens_threshold: 1000`?
Thanks for that, I used the option which designates a deep search and it improved. I’ll compare when I’m out of the garden
These options helped (in different ways). We are still adapting how our data is being seen by typesense. Is there a solution out there in userland where you can build a list of terms which would automatically filter out certain results?
Yup, you want to look at the curation feature:
There’s a lot of control there! Can search query use multiple words per rule, such as “chair or sofa” would filter out content which is not categorized as “seating”
It supports multiple words, but as literal strings, not as boolean operators
For your example, you want to create multiple rules
Indexed 3051 threads
Lightning-fast, open source search engine for everyone | Knowledge Base powered by Struct.AI