17 comments

[ 3.2 ms ] story [ 52.9 ms ] thread
So not just index the page while crawling but also some NLP? Good luck!! For direct results I like Wolframalpha.
Trying to be more like Wolfram imo, there's still lots of searches where Google fails for me, for example doing a search for "est time" currently gives the time in Eastern Kenya. Wolfram gives me the EST time in Australia (because that's where I am), if I change the query slightly to "east usa time" then in Google I get nothing, in Wolfram they show me the difference between the EST time in Australia vs the USA.
I'm in Sydney right now, logged into my main account, and [est time] gives highly relevant results for Eastern Standard Time. I get the same results in an incognito window. In fact, for every query you gave, I have the same results, though [east usa time] also gives the current time in East, Iowa.
I am from India and I get "6:37 Thursday (CDT) - Time in East, IA, USA" as a result. I think there is some component of personalization that might be affecting your search results.
And that's how google broke search in my eyes.
i also noticed this in the past few weeks. google now tells you what your IP address is for "my ip", sweet!
"Catch up to Siri in mobile search"? That must be some kind of hidden advertisement/typo/fanboyism.

Unfortunately though, there is definitely something wrong with Google results ever since the Panda update. It is very difficult to get any meaningful technical results now. I feel they had to dumb down search to be able to serve up a bigger variety of results.

If possible, could you provide example queries with poor results? The quality teams are always looking to improve results, and examples of queries with poor results are the quickest way to improvement. Or, at the very least, could you narrow down "technical results" to a particular field of inquiry?
For several weeks (months?) it would "correct" my search for pyquery and change it to jquery. It was quite annoying as I had to search for it many times.

It's fixed now, but Google's automated algorithms now are overly aggressive and assume I'm an idiot. For example, I have to +"<term>" a lot or it just ignores my term. I'm sure you can dig those up in your logs. Sometimes even that doesn't work.

Someone was scanning for a crazy backdoor, using this: /?dgd=1. Try searching for /?dgd=1 or "/?dgd=1" (with or without verbatim on).

You are going to get results about:

  dgd-1
  dgd 1
  d g d 1
  dgd#1
  dgd (1
I figured most folks knew that Google doesn't support arbitrary punctuation/strings. Do any search engines? I tried bing and duckduckgo and got nothing.

Don't get me wrong, I'm sure the quality folks would love to be able to support searching the web for arbitrary strings... but that's not something any search engine can provide right now, is it? It's orders of magnitude harder than current web search offerings.

Google Code Search supported [1] regular expressions. Google Web Search can match punctuation in some cases, but it seems to do best when the symbols are at the beginning or end of the search term. Examples:

http://www.google.com/search?q=%22c%2B%2B%22

http://www.google.com/search?q=%24100001..%24199999

[1] Actually, Google Code Search is still around. http://code.google.com/codesearch

Google Code Search does not (and never did) search the entirety of the web. Its index is orders of magnitude smaller.
By technical, I meant in my own field of web development/mobile development, but I'm sure it applies to other fields as well. I can give a few general examples of where I've noticed that results are worse:

- Stack overflow and other forum-type pages don't show up with relevant content, though if you dig deep enough through the pages you will find it. It used to be that I never needed to go past page 2.

- As a corollary, there are sites like bigresource.com that show up all the time and they only exist to scrape content from other sites, and these guys show up near the top. I thought that this was a big reason for the update in the first place, but I only noticed them after the update. I also can't seem to block them when I hit the back button (this only shows up sometimes)

- Way more aggressive auto-correction. This should be an option that is adjustable. There are many technical terms that are spelled a certain way that the system attempts to auto-correct and if I go with verbatim, it seems that this hurts results quite badly. Maybe the indexing itself does some auto-correction?

I don't remember specifics, but the most recent one was trying to find information on SIP and SIP trunking - it's taken me 2 weeks and I'm sure it would take me less if I had better results (like I had in the past).

All is not gloomy though. I love that you can now use some special characters: this has saved me many times.

Until a few weeks ago when it shutdown, I used Scroogle to proxy my searches, and it gave great results. 50 plain links without any other distracting content. Of course, they simply queried google servers with an IE6 UA, but it still allowed for all the search operators. I had been doing that for the last 6 years, and I must say I am at loss these days. I use DDG now, and it seems adequate...
Using DDG, quite often I don't even have to click any search results to find what I was looking for (usually a quick documentation snippet).

I feel that in old times I had that with google too, nowadays I often have to scroll around and even go to next pages looking for it.

I am surprised no one brought up the most important issue: Google are transitioning from content delivery to content production. This is a huge shift and for me shows the first signs of trouble; content producers will be a much lower-quality job at Google, and it will need lots of people to do that. It feels like Google is slowly becoming Microsoft.