Is there any way to use TFIDF along with ADA?

API
1

Is there a way to use TFIDF vectorization and ADA?
I am noticing with just ADA some words are weighted too heavily even though it is just a common title name like the company name where the documents originated from.
For example a questions like “What is product Thing from Company A” perform much worse in document retrieval compared to “What is product Thing”. Too many chunks being retrieved end up being just about Company A rather than about the main topic at hand. Which is the product.

2

Yes.

This article has the information that you need. The important bit is the RRF

Here’s a more hands-on article.

5 Likes