3 ms·

I think that double-quotes has had its Eternal September. Maybe "regular people" used it for mundane search tasks and didn't find what they wanted, Google detec

I think that double-quotes has had its Eternal September. Maybe "regular people" used it for mundane search tasks and didn't find what they wanted, Google detected this through their search metrics, and now they've decided to nudge double-quote search terms towards synonyms and typo-corrections... verbatim no longer. For the majority of seekers, Google may actually be performing better, but for a minority of hobbyists/specialists/experts looking for niche and jargon terms, its anything but an improvement.

eitland 5y ago
I've thought so too but it goes deeper. If this was the entire reason they shouldn't have botched the verbatim option as well. My guess is incompetence or arrogance or something. I don't mean incompetence as dumb and I'm not saying I'm smarter, but clearly something is missing that the previous maintainers had. My best guess is they threw out a bunch of integration tests somewhere around 2006 and never recovered ;-)
zbuf 5y ago
Presumably there's a technical challenge with offering to match exact strings at this scale? After all, the ability to do it was Google's main improvement over AltaVista at the time. I'd love for someone in-the-know to share some of the technical details. Because I could bet that the sloppyness with quoted strings could be about dumbing down, but could equally be a consequence of some implementation choice at this scale and the way the data is indexed.
- marto1 5y ago
  My humble guess is you'll never get these sorts of technical details from an internal so don't get your hopes up :-)
  - zbuf 5y ago
    If you don't ask :) I've never worked on a large-scale text indexing so it doesn't have to be a particularly sophisticated insight. I'm speculating that the majority of uses cases need some kind of work equivalency, so the major index is built that way. Whether that's just for the search terms of the average user, or ranking/clustering used to derive which pages are important. Then any additional indexes is a trade off with the magnitude of resources it would consume.

eitland 5y ago

I've thought so too but it goes deeper. If this was the entire reason they shouldn't have botched the verbatim option as well. My guess is incompetence or arrogance or something. I don't mean incompetence as dumb and I'm not saying I'm smarter, but clearly something is missing that the previous maintainers had. My best guess is they threw out a bunch of integration tests somewhere around 2006 and never recovered ;-)

zbuf 5y ago

Presumably there's a technical challenge with offering to match exact strings at this scale? After all, the ability to do it was Google's main improvement over AltaVista at the time. I'd love for someone in-the-know to share some of the technical details. Because I could bet that the sloppyness with quoted strings could be about dumbing down, but could equally be a consequence of some implementation choice at this scale and the way the data is indexed.

marto1 5y ago
My humble guess is you'll never get these sorts of technical details from an internal so don't get your hopes up :-)
- zbuf 5y ago
  If you don't ask :) I've never worked on a large-scale text indexing so it doesn't have to be a particularly sophisticated insight. I'm speculating that the majority of uses cases need some kind of work equivalency, so the major index is built that way. Whether that's just for the search terms of the average user, or ranking/clustering used to derive which pages are important. Then any additional indexes is a trade off with the magnitude of resources it would consume.