3 ms·

gitmemory or these translated stackeroverflow clones are a pain - the original result doesn't even appear anywhere in the results - it's the same in german. Ov

by nisa 5y ago

gitmemory or these translated stackeroverflow clones are a pain - the original result doesn't even appear anywhere in the results - it's the same in german. Overall google result quality declined a lot from my perspective (and colleagues also agree) in the past years. the info-boxes and sidebars are often just plain wrong (wrong product, wrong answers) - have to add quotes to every other query but these also don't seem to work - websites that used to be indexed disappear and I get lot's of noise and outright spam websites with redirection spam (I've tested on Linux with different browsers, unlikely that it's hijacked). Not sure what's going on - I guess they made a lot of progress in NLP with BERT and other models and nobody internally want to see the drawbacks?

Akronymus 5y ago
> it's the same in german. reminds me of msdn, where I always search in english through some search engine, then get the german version presented with a SMALL button to turn it to english.
lolinder 5y ago
Gitmemory et al are awful. They've made me think seriously about the plausibility of building a custom search engine. From what I've heard, the biggest problem with building a general-purpose search engine seems to be the index—indexing the entire web requires data centers many times larger than most can afford, and deciding what is relevant in such an index requires excessive training data. But while working, I typically only want my search engines to search the same 200-odd sites (GitHub issues, Stack Overflow, MDN, language docs, high-quality technical blogs, etc.) It's too many to know which "site:" query to use, but it's a small enough set that it should be a manageable problem for an upstart search engine. It also seems like this is likely true in many domains. Taking on Google directly by trying to build a general purpose search engine seems to be a non-starter (most competitors just crib off of an existing player, and don't do much better). But a bunch of sharded search engines for specific domains could theoretically outperform Google/DDG/Bing within their respective domains, because the human using them has already narrowed the domain down dramatically. They wouldn't dethrone Google (general search is always going to be needed) but they could compete for a segment of traffic.

by nisa 5y ago

Akronymus 5y ago

> it's the same in german. reminds me of msdn, where I always search in english through some search engine, then get the german version presented with a SMALL button to turn it to english.

lolinder 5y ago

Gitmemory et al are awful. They've made me think seriously about the plausibility of building a custom search engine. From what I've heard, the biggest problem with building a general-purpose search engine seems to be the index—indexing the entire web requires data centers many times larger than most can afford, and deciding what is relevant in such an index requires excessive training data. But while working, I typically only want my search engines to search the same 200-odd sites (GitHub issues, Stack Overflow, MDN, language docs, high-quality technical blogs, etc.) It's too many to know which "site:" query to use, but it's a small enough set that it should be a manageable problem for an upstart search engine. It also seems like this is likely true in many domains. Taking on Google directly by trying to build a general purpose search engine seems to be a non-starter (most competitors just crib off of an existing player, and don't do much better). But a bunch of sharded search engines for specific domains could theoretically outperform Google/DDG/Bing within their respective domains, because the human using them has already narrowed the domain down dramatically. They wouldn't dethrone Google (general search is always going to be needed) but they could compete for a segment of traffic.