Ask HN: do you feel Google search result quality has gone down?
The thing is, I'm not sure if it's because I do a lot of very specialized stuff these days, or because the search quality really has gone down. Consider these two examples:
Search for "Linux asynchronous IO". You'll get a lot of articles, but most are four years old (which is an eternity in the Linux world). These results aren't very good - posix AIO is implemented in userspace threads, and io_submit and friends don't work in many cases. Which cases? Hard to tell - I couldn't find any information in the results no matter how long I searched. I couldn't find any benchmarks either.
Perhaps it's because there is no good info on this on the web (hard to believe). So let's try something else - search for "concurrent hashmap in C". After hours of searching and playing with keywords, I got almost no useful results (other than Intel's libs, but not too much info on that either). It's difficult to believe that there are no good implementations out there.
So, is it the specialized nature of my searches, or is it Google? What do you think?
112 comments
[ 3.3 ms ] story [ 101 ms ] threadLast night I was searching for the syntax of DEFTYPE when used with various types (i.e. MEMBER, SATISFIES, OR, etc.) and the #1 his for "deftype member" was the personal MySpace page of some guy.
I think they're optimizing for "social" results now.
And, yes, the results are different. I'd agree with you they seem enhanced (eg., classified results) and fresher for popular culture, but somewhat worse for domain specific queries.
It would probably be a highly praised feature to seperate off a second index like that, as specifically searching for programming language concepts and documentation can be difficult (the C# and .NET problem).
One thing I noticed is that searches no longer require that all words in the query be present in the search results. Adding a + before a word is now required to ensure that it's present in results. That frequently results in me having to do 2-3 searches to find something that could previously be found with one.
User testing must have somehow played a role, I hate to blame non-technical users but... still I can't believe anyone gets better results when the keywords are optional.
The odd thing is I remember switching from AltaVista to google, before google you always had to discount the first bunch of results, but google was just so amazingly accurate. And yet now I find myself skipping the top results in a google search.
I remember when google bombing first started, it didn't bother me much, but then google tried to counter it and I could swear searches got a bit worse. And recently they've gotten even worse.
It's a shame this seems to be destiny of all truly great things.
Plus, google's handling of punctuation (e.g. f-script) is a pain since (even with the +) it will do weird substitutions and consider blanks good enough.
This seems to be an optimization for their average user, but is really inconvenient for people searching for system errors, mathematical/cs theory terms, or other queries where acronyms are common.
However, I do think it could be possible that the pagerank algorithm does not scales good enough for the growth of Web pages, and especially with the "SEOs" out there always trying to come up with new ways to game the system.
On the more important metric of 'has quality gotten worse in the last couple of years', I would say 'sort of'. The direct quality of results HAS suffered, but on the other hand, google have implemented the user-wiki thing that allows you to modify, to a degree, which sites are less relevant.
I will add that I think google needs to rethink it's keyword fuzziness, in the past it used to be acceptable if the results didn't exactly match what you were searching for, but these days that is becoming more of a problem. If I search for a bunch of words, I typically know I want those words, by all means suggest 'did you mean ... ?' but the fuzziness in the results needs to be pulled back.
Let me guess:
1. Google's Pagerank patent is about to expire, and they have to make something new?
2. New ranking algorithm adjustment to compete with Bing but failed?
And similar discussions on reddit:
http://www.reddit.com/r/programming/comments/9qro1/
http://www.reddit.com/r/programming/comments/9s6p2/
They should drop yahoo answers, any site with an affiliate link, and any of the internet marketer sites ( like ezinearticles.com ) from their index
My friend had an idea to make search rankings go by the number of ads on a page, the less ads the higher the ranking. I guess Google would never do that because of their adsense program.
Edit: redundancy
And it's duckduckgo.com that helped me.
?
Although I alternate search engines regularly, I do think the way Google indexes for specialized searches is pretty smart.
Only 5 topics, unfortunately.
Thanks for the link!
http://www.google.com/#hl=en&source=hp&q=site%3Avldt... http://www.bing.com/search?q=site%3Avldtr.com
As long as the Google has the most up to date index there is no good alternative for me.
You mention trying Yahoo. Did it give you better results?
See also here: http://news.ycombinator.com/item?id=464393
Another example: when I search for "vldtr is now a couple of months old" (from my post up there) Google returns this thread, Bing returns nothing.
Different metric, same result.
SEO: When you cut through all the BS, the entire goal here is to make a less good match come first. And it works (sorta). Just consider crap sites like Experts Exchange that we've only learned about because they pollute many searches.
Feedback effect: Thanks to google, less people do less collecting of good links. Why bother when you can google for it? So there's less good information for google to use in ranking links. Bear in mind that when google started, nearly every home page had a long list of links to all the pages that particular user liked and frequently used. I used to have one; I've long since deleted it; my blog has some outgoing links that I like, but relatively few. If I twittered, I'd probably post a lot of outgoing links, but of dubious value; there's no gardening of just the perfect page of 100 links going on anymore.
(I think this also partially explains why some (generally more specialized, so less effected by other things) results feel dated -- legacy links that are still hanging around from days when links were still used that way.)
Feedback effect: Thanks to google, ten sites tend to be more important than any other sites on any given topic. This results in certain sites becoming increasingly important. Wikipedia is the chief example here. Why is there only one Wikipedia and not a dozen? Chiefly because it's gotten all the google juice. If you want your wiki article on foo to show up in google, you naturally write it on Wikipedia, not Fooipedia. The result here is that all google searches feel increasingly the same -- of course Wikipedia is always in the top ten, or maybe something like Stack Overflow for a technical search.
----
So, these days, if I don't see something interesting in the top ten, I often click on the link to page 10 (or 20, or 100) of the results. Often more interesting. For example, google for "mashed potatos".
Top 10 results: "Perfect mashed potatoes" (SEO), allrecipies.com (always in top 10 for any recipe search), foodnetwork.com, Wikipedia, about.com, nytimes, etc. Pictures of mashed potatos. All generic and useless.
Page ten results: Dairy-free mashed potatoes. _Potato_ free mashed potatos! Caramelized Onion Horseradish Red Mashed Potatoes! A poem about eating them. At least marginally more interesting and quirky. What I would have expected out of google circa 1997.
I have started tailoring my searches in odd ways to help them out. Ex: Adding a the year when I want current results. But, without useful links it's all GIGO.
A related point: whoever first owns the data from all link aggregators (digg, reddit, mixx, etc) and all URL shorteners (bit.ly, tinyurl, ad nauseum), and weighs those results more heavily is going to have an awesome search engine... albeit better for entertainment than productivity.
Except I'd never find the obscure stuff that Google helps me find every day. I don't think a lot of links to manuals,mailing lists, etc show up on any of the sources you mention.
Google already does this, by surfacing youtube videos in many search queries.
i will humbly disagree. i think folks who browse the web are different from those who search the web. search is what gets you the most relevant results, therefore more opportunity for ad money.
Until they implement it, you can use the CustomizeGoogle add-on for Firefox that lets you filter results out on your side.
Now, if you want something "quirky", why are you searching for a generic term? What kind of "useful" result do you want from a search on mashed potatoes? If you give them a crappy search query, they should be giving you as generic of results as possible.
One thing I've found is that if you are looking for something specific, don't search for something generic. If you wanted something "quirky", why didn't you do "mashed potatoes quirky"? Then you get a restaurant that features mashed potatoes heavily in their recipes, a carmelized onion mashed potato recipe, a mashed potatoes festival, several more "interesting" recipes, and a book called "Grinning in His Mashed Potatoes".
It sounds to me like the results have improved, not gotten worse, if you aren't getting a poem about mashed potatoes on the first page of search results for just "mashed potatoes".
That is not the entire purpose of SEO. There's good sites out there that don't provide their content in a way that can be indexed by spiders. SEO often solves that. There certainly are plenty of people making bad websites and trying to make them rank, it's Google's job to weed out the useless information.
People still collect links (shameless self promotion: http://internetmindmap.com ), they still have huge blogrolls, there are human powered search engines, a vast amount of directories for every imaginable niche...
Google doesn't need the perfect page of 100 links and I doubt it ever did.
Your mashed potatos example does not make sense. Google gave you generic info for your generic search query. How is that bad?
Now the fact that certain sites dominate a very wide range of search queries, is an interesting point. Personally, I would just add a sidebar or something similar, to be occupied by the "staple" sites, such as wikipedia, about.com etc.
I think in future search engines should lean a particular way depending on your profession - "Google for Medical Personell", "Google for Architects", "Google for Programmers".
Strange that with regard to current events the web seems to have little historical memory, and yet with regard to current technology it has too much.
http://news.google.com/archivesearch?q=2002+house+vote+socia...
on a similar note - http://guillaume-nargeot.blogspot.com/2009/08/think-twice-be... (probably saw this on HN)
It does work as described! - which may very well not be as desired. If you want a more mathmatical-logical use of operators then you need to go find another search engine.
http://www.google.com/support/forum/p/Web+Search/thread?tid=...
I am afraid that Google is now more of a 'social search engine' than a 'hacker search engine'.
FWIW, Bing nailed it on the first query.
Seriously, the problem is advanced users get unexpected results with the query expansion and refinement layers they've added on. While dropping obscure words is helpful when grandpa has misformed queries, it's maddening for a technical user looking for a very specific, infrequent keyword. However for grandpa, it's probably a better experience for a generalized result.
Using +'s works well enough, but it's disappointing we have to use a less efficient method of querying now.
Google was basically filled with dead end forum postings and SEO spam.
DuckDuckGo was more helpful and brought me to the MS TechNet article with full documentation on MS Communicator Policy configuration.
Bing was surprisingly the most helpful and brought me to the Communicator Team posting from 2008 which shows me where I should have been able to find the logs.
It looks like my work has blocked/disabled this feature on a global setting even though I haven it enabled locally.
Here is a search that I have had problems with :
octave "--eval"
Your site does pretty well with this (fourth link is somewhat relevant).
Previously it was equivalent to ("hyphenated word" OR +hyphenatedword). But now it seems to behave almost the same as the unquoted (hyphenated word).
Just to make matter worse, when I tried out my example just now I found that the first result (a wikipedia page) for "hyphenated word" doesn't even include the phrase!
Notably, you return a result for a specific osCommerce error message that I wrote about a while back (!); Google doesn't even know I exist.
You might try to track page age; some of the results I get are from 5 years ago and as noted by others, aren't always useful today. But that's a harder problem for another time.