582 karma · joined May 2, 2023
Roughly, first there is the query analysis/manipulation phase where you might have NER, spell check, query expansion/relaxation etc
Then there is the selection phase, where you retrieve all items that are relevant. Sometimes people will bring in results from both text and vector based indices. Perhaps and additional layer to group results
Then finally you have the reranking layer using a cross encoder model which might even have some personalisation in the mix
Also, with vector search you might not need query expansion necessarily since semantic similarity does loose association. But every domain is unique and there’s only one way to find out
This is important because libraries change, introduce new functionality, deprecate methods and rename things all the time, e.g. Polars.
> study including 22.7 million vaccinated individuals and 5.9 million unvaccinated individuals
These are the important bits for the non medical folks
This is an absolutely wild (and incorrect) thing to assume. The problem of content lock-in is anti-competitive and it would be better solved without mergers
This bit got me. It's a direct quote from the linked post for those who haven't read it
To stick to the metaphor (apologies if this isn't HN friendly)
Smart TV? Fart TV
Smart display? Fart display
Smart fridge? Fart fridge
Keyvi is also interesting in this regard
I wonder why we don't see more software engineering co ops?
People use web browsers to hit websites, but when code hits URLs they are typically just called APIs. A website is technically an API too
"What if we looked at advertising, but not for too long?"
I'm interested in how to bake accountability into an organization. I don't like the idea of using whistleblowing as a crutch because things have to get really bad for someone to blow the whistle.