News for hackers. Stuff that matters.
Re:Search engine startup from former Google Ads boss raises ... by dstick on Thursday March 11, 2021 @02:47AM (#26421412)
Yeah, you do need ML. But a tweakable one. I’ve built a crawler, scraped the first 10 pages of all public domains I could find, then made those billions of pages freely queryable with ElasticSearch. The result: moderately good. It was great for discovering hidden gems. But you needed a lot of longtail keywords and exclusions.

There is a massive amount of textual garbage out there!

Re:Search engine startup from former Google Ads boss raises ... by axaxs on Thursday March 11, 2021 @03:01AM (#26421492)
I gave it more thought in passing and am a bit curious what real, heavy ML might look like. And if there's value in such things.

Example: I often find myself remembering bits of things, and querying Google, and the results are way off or completely irrelevent. Things a human would know in a matter of seconds, generally.

I was for example trying to remember that newish Netflix movie that is very similar to the Michael Douglas movie "The Game." It feels like such a question should be easy for a sufficiently smart system. Something like Watson, I guess?, at the search engine level would be immensely mindblowing to me.

The answer btw for anyone curious was "Rebirth." Nearly every review of said movie makes the comparison.

Re:Search engine startup from former Google Ads boss raises ... by rzzzt on Thursday March 11, 2021 @03:42AM (#26421749)
This is where semantic data would be really handy (admittedly, it's not a great example because the two movies are not linked at the moment):

- https://dbpedia.org/page/The_Game_(1997_film)

- https://dbpedia.org/page/Rebirth_(2016_film)