Neural Search Talks — Zeta Alpha cover image

ColBERT + ColBERTv2: late interaction at a reasonable inference cost

Neural Search Talks — Zeta Alpha

00:00

The Importance of Similarity Scores in Query Models

There's no pre-processing applied to the terms that you search through. They have any sort of intuition on the qualitative behavior of what the scores of coders actually are doing. It might be that this model is very sensitive to the vocabulary size, right? That it might become much better, much worse if you instead of like 30,000 words, you take like 10,000 or 60, 100,000 or...It's possible, yeah. I think there's been very limited experimentation of changing the vocabulary just because it's really painful to do. You have to train bird from scratch, which is pretty useful.

Transcript
Play full episode

Remember Everything You Learn from Podcasts

Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.
App store bannerPlay store banner