Here’s the thing though. You absolutely could still use operators reasonably well even if the results are fuzzier.
You just use them to control how you leverage the algorithm. AND feeds the algorithm the two sides and filters to results that appear on both. OR joins the two result sets. “Filetype” filters the result set for results that are the relevant file type. Etc.
If they’re not that common they’re not going to have meaningful costs, especially when most power users don’t use them for most of their searches.
Thorry84@feddit.nl 3 months ago
Most search engines have filters for stuff like filetype, limiting the search to a specific site and filters for time and location (when applicable).
Like I said, search engines are way more complicated than one might think. Doing the kinds of things you mention would be hard and only very few people would need something like that. There are tools out there that do meta searching for analysis though, so you can use search results as data in your analysis. Most of those are highly specialized and often paid, but when you need them it’s worth the price.
Remember companies like Google invest millions (if not billions) into their search engine and have huge teams working on them. Anytime someone says: “Why don’t they just…” the answer is probably very long and complicated.
conciselyverbose@sh.itjust.works 3 months ago
No, it isn’t even a little hard. It’s super simple pre-parsing of the input that can trivially be done client side before the query even touches their server. Advanced users who use those tools are perfectly capable of taking the extra step to indicate to the engine that they’re doing a real search, and the worst case is still far less intensive per search than any of the LLM nonsense they added to every search and is almost never useful in any way.