spiral geyser
Let's say I turn a whole page of text into a single vector. This page includes different kinds of information, e.g. my passwords, birthdates, book titles, all kinds of stuff.
Now I create a query about one single thing, .e.g.: "What is my WiFi password?"
Will the large page vector be positioned close to my query's vector, even tho the page contains a bunch of different information? Or do I need to slice the large page up into several separate vectors to get good results?
plucky light
If somewhere in your vector the tokens for password appears, it’s likely that a query for “what’s the password” will indeed return valid results.
If you’re working with pinecone for example I’d return maybe the first 3 closest results, just to be sure.
However, the word “password” in such context lets my skin go chicken 😅
Please don’t. I’m assuming you just used this as a reference to illustrate the case?
spiral geyser
If somewhere in your vector the tokens for password appears, it’s likely that a ...
Yea, it was just meant as an example. I'm not actually storing passwords. 😅 Thank you for the explanation.