|
|
|
|
|
|
by logicprog
866 days ago
|
|
|
I really think the existence of this project is a Good Thing. Massive kudos to the people working on it. Previously I was always disappointed that our only options seemed to be open source meta search engines and closed source search engines, with most of the latter being corporate surveillance calitalist hellscapes, or anonymized portals to the same (Google and Bing, Duck Duck Go and Startpage), with only Kagi being an exception, although still closed source. It seemed like a really uniquely bad landscape, given that in most other areas of software there are at least some FLOSS alternatives to proprietary platforms that actually implement their core functionality, regardless of their relative quality. Stract finally changes that, which means several good things to me. First, you get actual, real accountability for them to stick to their stated privacy goals. Second, you get the ability for a wide variety of people to contribute to and influence the project, and/or learn from the project how to do this stuff themselves. And finally, most importantly, since it's a full reimplementation from indexing up, it's an opportunity to innovate on and experiment with the fundamentals, instead of just rearranging deck chairs on the titanic like e.g. Startpage. Thats really great :) |
|
|
- search results seem to be somewhat case sensitive, which is a massive problem for me when for instance searching programming terms
- in general the matching algorithm seems way too strict, only matching against the exact thing you entered, which makes it very difficult to profitably search for things where you don't have exact specific terms in mind, like perhaps computer errors or genres of things. I think a lot of people's problem with how liberally Google interprets your search results is not that it interprets them liberally necessarily, but that it doesn't respect the other options it provides for trying to match things more strictly. As long as you provide something like Google's quote mechanism and actually respect it, I feel like it would be a lot better to match things more liberally by default. Maybe some amount of fuzzy searching, and matching by synonyms. Also you could probably just use a dictionary of synonyms to do that instead of whatever statistical model Google is probably using, in order to ensure more predictable results.
- as someone else somewhere in this thread mentioned, it seems like stuff like a, an, and the, are all matched against and waited equally to other words. This, especially combined with the fact that it only matches words exactly makes the search results feel way too brittle and unforgiving