Combining sparse (full-text/TF-IDF style) search with dense semantic embeddings improved citation retrieval accuracy over using dense embeddings alone, because legal terminology and exact citation wording benefit from exact-term matching that dense embeddings can miss.
Alex describes how adding sparse retrieval (via Pinecone's TF-IDF-like vocabulary index) back alongside dense embeddings measurably increased citation accuracy in Sphere's evals. ✦ AI generated
Alex Boucott · The TWIML AI Podcast · 2026-06-09 · original ↗
starts at this moment · 22:17
“Can you dig into a little bit more detail on the the dense versus sparse um aspect of what you're doing?”
There are times especially with certain when it comes to citations um and and pulling out certain terms from passages that come from the dense embeddings where you also want to search sparse where you want to do you know a full text search of certain uh words and certain terms and pull those in as well... and what we saw was a pretty good increase in accuracy on the citation side.
verbatim transcript · starts at 22:17
22:17dense embedding, right? That would that semantically embed these uh passages um and so that that is what we started with. I think what we found and when we brought sparse back into it was there are times especially with certain when it comes to citations um and and pulling out certain terms from passages that come from the dense embeddings where you also want to search sparse where you want to do you know a full
22:45text search of certain uh words and certain terms and pull those in as well so that you can and then and then compare the two of them and what we saw was a pretty good increase in accuracy on the citation side. So we have some evals that we run on the retrieval part that um we have a baseline of like these citations right these passages are the
23:08ones that should be retrieved for these queries and as we kind of layered sparse back into that we saw another we saw an increase in accuracy and so that's that's kind of what we stuck with. >> By when you refer to dense and sparse are you it sounds like you're talking about embeddings versus full text search as opposed to like two tiers of embeddings or something like that.
23:35>> Yeah that's right. So yeah dense dense is definitely embeddings like yeah semantic embeddings that we use open AI's embedding models for and then yeah when I say sparse I'm referring to in our case we use pine cone to essentially create a sparse representation so you we've loaded a vocabulary and then each passage is fed through and it keeps an index like full text search of the different terms and their
24:03different usages across passages. So it's not quite you know elastic search or Apache Lucene but it's it's sparse and like a TF-IDF type implementation. >> Got it. So you've got a predefined vocabulary and as you pass these documents in the pine cone it's just flagging which documents talk about which of these terms. >> Yeah which passages are talking about which terms so then you can search so that when then you can search over