Hacker News Reader

Comment by iamgopal | Hacker News Reader

Am I the only one who is seeing this sudden surge of bbc.com linked articles ? Not that I'm complaining.

When people started making pages for Search Engine instead of Human Consumptions, Internet died. "Light on information and Heavy on keywords and linking" sites and pages flooded the internet, through which finding the actual content, needs another level of skill.

I think solution to this is, an AI based open source browser extension, which find relevant page from search engine result depending upon its usefulness index. ( or just show its usefulness index, along side the result, in short sabotaging the Google search results ;) ). Is there anything similar available ?

View on HN

It's an Eternal September problem, not a search engine problem. Whenever I use search like a proper normie to buy something, check song lyrics, or weather, the results are excellent.

On that note, lately it seems Google results are more and more about what’s ”popular” in some sense instead of what’s relevant, especially when your search term has multiple meanings. Two recent examples of things I’ve run into are ”cloud” (pages and pages of cloud computing results, no mention about the meteorological phenomenon) and ”thrush” (nothing about the songbird family, just dozens of results about the yeast infection). Whether I’m logged in or use private browsing doesn’t make much difference. Not sure what has changed, but I’m used to Google giving a more representative sample of different meanings of a search term.

Duck Duck Go has a similar problem for cloud and thrush, but if you click the “meanings” tab, it lets you switch results over to what you meant, and then gives decent results.

More broadly, DDG implements UI features that allow users to clarify their intent, while Google focuses on using machine learning to infer intent.

When things go sideways, this means DDG empowers users, and google overrides them.

Disclaimer: I switched away from google search years ago, and now I find its user interface completely baffling.

Maybe I just don’t know how to use it any more, and there is some override I’ve overlooked.

Also, I’m kind of shocked by how many corporate logos and spam links fly by when I scroll down the Google results for “cloud”. It feels like the 90’s internet before animated gifs and the blink tag were invented.

In fairness to google, ddg has a similar number of logos, but they’re each approximately the size of one normal sized character, not one third the width of the screen like in google.

> Duck Duck Go has a similar problem for cloud and thrush, but if you click the “meanings” tab, it lets you switch results over to what you meant, and then gives decent results.

This is a good idea but not well implemented. For example, searching Python displays a relevant carousel of "meanings" - you have the snake represented, and the programming language. But choosing an item from the list simply changes the search term.

This is particularly bad with "Ruby". Perform a search for Ruby and then click the 'a pink to blood-red colored gemstone' meaning. Observe that the top result now that you have clarified is "Ruby Programming Language".

It's even worse than that - it also appears to learn synonyms from popular usage, even when it is incorrect or misleading.

For example, when it comes to firearms, there are two completely different parts called "extractor" and "ejector", that people often confuse because of similar sounding names. So now, when you search for "AR extractor removal", the first hit on Google is about ejectors; and it specifically highlights the word "ejector" as bold, as if it came from the search query:

https://www.google.com/search?q=AR+extractor+removal

But are relevance and popularity not related? If more people are using Google to learn about cloud computing than the clouds in the sky, I’d hazard that that would make cloud computing the more relevant result, in terms of what Google sees anyway.

It’s tricky, though, because I’d say that the clouds in the sky are more relevant to humanity as a whole. Not everyone cares about or needs to think about cloud computing.

Both are relevant meanings and should be included in top results. Perhaps Google should have a disambiguation feature like Wikipedia does (and DuckDuckGo does according to another commenter).

They used to - I remember pushing this feature hard when working on the search visual redesign in 2010, and getting the response that it was just a really hard engineering problem. The back-end teams delivered sometime in late 2011, and IIRC it was live in the left-nav (remember when Google had a left-nav?) for a few months to a year.

It was unlaunched eventually because nobody ever clicked on it. It turns out that disambiguation is a relatively rare feature for people to want when most of the query stream is navigational queries, and it's contrary to Google's goal of getting you to your answer as quick as possible (2 clicks if it were a UI feature vs. one click if the result were just in the results). AFAIK the backend is still live, but it triggers off keywords in the query rather than any UI element. If you search for [cloud weather] 100% of the results are about the meteorological phenomena, if you search for [thrush bird] 100% of the results are about the songbird, if you search for [jaguar car] 100% of the results are about the car, and if you search for [python snake] 100% of the results are about the reptile.

Interesting, thanks! Yeah, disambiguating with a more specific search query is, of course, a fine solution. I guess I've just became accustomed to Google basically reading my mind compared to the bad old days of search engines when figuring out the proper incantation to get the results you want was akin to a logic puzzle.

I absolutely agree with you there. In the meantime, adding additional descriptors yields better results but is maybe a bit un-intuitive to some users (e.g. "thrush bird" rather than "thrush").

The solution is they need to bring back webrings.

Federated search (please don't say "AI") may eventually become part of the internet. It's garbage now, and the skills required to make it go are generally employed at google, bing or yandex, but it's one of those things that should have happened.

> Am I the only one who is seeing this sudden surge of bbc.com linked articles ? Not that I'm complaining.

I think it comes in waves. There was a surge of TechCrunch articles the other day. I noticed because they're blacklisted in pi-hole due to their crappy UI.

Reply on news.ycombinator.com