Ask HN: Which websites actually have a useful search that Google can't beat?
Everybody uses Google, Bing, DDG, or some other search engine. But are the search giants as ubiquitous as they seem? I think there's a great value to individual sites' search features, mainly because they go beyond the grasp of GoogleBot.
There was a post recently about Google's main competition being Bing. I think that's entirely untrue. I think their main competition is the long-tail of search that they simply don't have the dataset to compete against.
Here are sites that I can recall using the search feature, and it actually being useful:
- Wikipedia (half the time I use Google, though)
- SearchYC
- Urban Dictionary
- BTJunkie
- StackOverflow
- Delicious
- Twitter
- YouTube
- eBay
66 comments
[ 3.1 ms ] story [ 128 ms ] threadIMDB can give me a list of all TV series / films / etc. with "Hobbit" in the title: http://www.imdb.com/find?s=tt&q=hobbit
Musicbrainz lets me search by track name, album name, artist name, etc.: http://musicbrainz.org/search/textsearch.html?type=track&...
I don't have any experience with their audio file (MP3, AAC, etc) fingerprinting, but I can tell you it's a great experience to pop in a freshly purchased CD and have MusicBrainz find it. I know that experience comes from some other kind soul having input the CD, so I try very hard to make sure I do the same.
I do, via the desktop program [Picard](http://musicbrainz.org/doc/PicardDownload). It really works well for properly tagging mp3 files (and renaming files to fit the tags) -- once they know the tracks. That often isn't the case with obscure, new releases. But you can't win them all. It's a program that's really worth using.
StackOverflow is a good example for this, I can usually reach what I was searching for in it with from a good google query.
Well, in Italy we have "pagine bianche": ahttp://www.paginebianche.it/index.html
Which is simply the translation of white pages: http://www.whitepages.com/
Regarding specialized pubs, its a long time I don't use one, but this list seems fairly complete: http://en.wikipedia.org/wiki/List_of_academic_databases_and_...
I've been working a lot in this space, and I think there needs to be a protocol implemented beyond OpenSearch, which is essentially limited to 1 text box, with some nice additions (i.e. you can specify a query suggestion URL as well).
There should be some sort of superset to OpenSearch. I'm suggesting an AbstractForm protocol which would allow websites to export their basic query functionality into an an XML document. Relevant providers (such as Ubiquity, and the thing I'm working on that will be launched soon), would be able to use this protocol to integrate the provider's AbstractForm in a convenient interface.
With Kayak, for example, AbstractForm might specify a "from" field of type "city", a "to" field of type "city", and some other optional fields (date, round trip, etc). You could then have a service which implements this however it chooses... perhaps a program with loads of AbstractForm interfaces loaded, and when you start typing "kayak" it knows to show you "To" and "From" fields, etc. I really believe this type of thing would be a major boon to the internet.
Google bought ITA and Bing bought Farecast for the purpose of getting direct access to this data. Since they can't web crawl this type of data, I would expect more acquisitions and/or licensing deals - such as the deal with Twitter.
Google has a decent foundation already with queries such as 'population of london' '<movie name>' (which shows local theatre times) and 'weather 94000' etc. Full list here:
http://www.google.com/landing/searchtips/#helpcenter
Flight to LA from where, when, how long, economy/first, what times of the day, etc.
Hotel room in New York next weekend which days, smoking or not, upscale or not, which part of new york, which new york?
Metallica san jose tickets are closer and could generally get you the venue's page that will then ask which showing, which seats, how many tickets together, etc.
eg. 'flight to LA' can show :
etc. etc.We've built resourcey.com trying to tackle this problem. That is, http://resourcey.com/site_details/2/news.ycombinator.com/ is the answer to "Am I missing something important about Hacker News?".
This is a different form of search queries we couldn't find a way to get results for via standard search engines. If you think this functionality can be somehow produced by a standard search query, do tell.
Right now, there are not too many metadata websites out there. Your site, AboutUs.org, Alexa and ilk, moreofit.com and ilk, and comment aggregation like BackType and UberVU, are all that come to mind when I think: "I have a URL, what can you tell me about it?"
if I need to find something on stackoverflow, I just go to google and do a site:stackoverflow.com
http://www.google.com/search?q=site:stackoverflow.com+%s
There is a fair bit of data on forums that is not exposed in the UI, simply because no frontend engineer has had a chance to look at it. I'm a frontend engineer, I'm potentially looking for a new project or some 20% work, and so examples of where Google isn't quite working right are pretty helpful to me.
Wolfram Alpha is phenomenal for certain types of queries. I find their results for companies pages to be so cleaner and more intuitive than either Google or Bing's.
I think there are huge openings for competing with Google and Bing when focused on a specific niche. Google's revenue per search query in the US is $0.12.[1] And, I know I"m getting frustrated with their search results, and I find myself using multiple search engines for different types of queries.
There's tons of room for competition in this market, and I don't know why more startups aren't taking advantage of it.
<shameless plug>
This is the reason that we're turning http://Newsley.com into a search engine for economic and financial news. The news sections of Yahoo Finance and Google Finance suck IMHO. We're trying to make financial news search suck less.
We're focusing on building our index and results pages like crazy right now. We're currently indexing the Economist, NYT Business and BBC Business[2]. Bloomberg News should be online this week.
After getting Bloomberg online, we're going to focus on getting the alpha version of our search released.
We're making our results pages available as soon as possible, however, so we can start building a bit of organic search traffic to our site even before search is released. So far, it's worked great. Since we started releasing those pages 6 weeks ago, our traffic has tripled.
In the meantime, each one of the keywords or tags below each article leads to the results page for that specific term. Feel free to click around.
</shameless plug>
ref:
[1] http://dondodge.typepad.com/the_next_big_thing/2007/05/why_1...
[2] http://newsley.com/crawl_stats
I'm finding myself going to DDG and Stack overflow for answers to those problems first, just because I don't have to sift through results from crappy forums or Experts Exchange to get to answers for my specific query.
I do prefer Google results to Bing's results for those topics, because I can query Google groups, which also tends to have higher signal::noise ratios.
As I said, Bing has Travel search down pretty cold as far as I'm concerned, and I travel quite a bit. That may well change since you guys bought ITA, which provides Bing's data. Brilliant move, btw. I do use http://matrix.itasoftware.com a ton after a quick overview search on Bing because of the domain specific language that itasoftware lets you use on your searches.
Amazon does much better for shopping and product search. My first inclination these days is to go to Amazon and look for a product. If the price seems reasonable, I have Amazon prime and the product gets shipped to me for free in 2 days, or I can spend $4 and get it shipped next day.
I found myself living over at Edmunds.com when I was searching for a used car to buy two months ago.
I think that the results from Wolfram Alpha on company financials are much better laid out than Google Finance or Yahoo Finance. I'm much more likely to use Wolfram than GOOG or YHOO to get an overview of a companies financials.
I think there's a lot that can be added when it comes to financial news search, which is why I'm building http://Newsley.com
Looking for recipes on Google pretty much sucks. I'd much rather go to the Food Network.
The reason that I think you guys are hurting at times in search results, is because your mission is to categorize "all the world's information". So, you're approach is to pull in as much information as possible, and then sift through and sort though what's important.
If a site takes a semi-supervised approach like DDG or provides search over a specific silo of curated content like Stack Overflow, then I think you guys are going to be hard pressed to compete.
And on a certain level, it doesn't make sense for you guys to try and compete with sites like Edmunds.com or Stack Overflow. You guys are looking for niches that can provide the next $0.5 Billion to $4 Billion in revenue. Small verticals like programmer searches, recipe search, financial news search etc... aren't really going to be worth a huge investment for you guys to dominate.
Speaking for myself, I'm not really all that concerned if Amazon or Kayak is better at their specific vertical than Google. That's why they exist, they can focus much more tightly on a good user experience for that particular task than a generalist search engine can, and that's basically the economy working correctly.
I do care when I can't do a particular task because Google's search results suck. That seems to happen more and more with programming-related queries, and I dunno if it's because I'm searching for harder, more narrow topics or if it's because Google's results are worse. I recently was looking for info on Haskell's FFI, so I searched for [CUInt]. The top result was "an Irish slang term for female genitalia". Not what I had in mind, Google.
Again, the most useful info we can have is the exact queries that went wrong. It's like any debugging problem: if you can't reproduce the bug, it's maddeningly difficult to fix it.
For some queries it's easy, like if I search for [amazon harry potter], I probably want either the Harry Potter Store or the Harry Potter books on Amazon.com, which happen to be the top two results. If I search for [llvm doxygen], I probably want the index page for LLVM's API docs, and that's what I get (even if it does have no snippet). If I search for [height of empire state building], I even get a little OneBox with the answer right on the results page, along with the sources it comes from.
But what if I'm searching for an obscure error message, and the answer is in a forum posting that was mirrored across a dozen mailing lists? Which is canonical? The biggest? The one on the site for the project I'm searching for? The one with the fewest ads?
Or what if I search for [can't read my poker face]? The #1 result is for metrolyrics.com, which sucks - it's a spammy lyrics aggregator with a popover ad. But at least it has the lyrics. Lady Gaga's official site doesn't.
It's those latter cases that interest me, but it's not exactly clear what the desired behavior is.
Next time I run into a search that would have a more appropriate best answer I'll try to post it here.
In terms of competing with Google, Bing, etc, I'm on it. I'm incredibly excited about launching my latest service, which hopes to gain a decent market share of start page users. Basically, it's going to be a search portal. Not another one of those crappy multi-tabbed "search 10 engines at once" pages, but rather something much more... substantial. That's why I started to thread, to see which sites I should include in the index. I'm glad somebody else sees the same potential in the search market as I do.
Craigslist and Facebook are both in the top 10 search engines.
I may be biased, since I wrote half of it. Funny seeing it turn into some kind of zombie :/
That's one of the first ways I will refine a search about a programming topic these days, but no need to use stackoverflow.com search.
To answer the question, I'd have to plug my own startup, http://historio.us. It's actually made bookmarking viable, for me.
http://www.shopwiki.com/LCD+TV+%3E43%22+%3C%241000?sb=1 (LCD TV >43" <$1000)
For example, here's the video card power search: http://www.newegg.com/Product/PowerSearch.aspx?N=100007709&#...
It allows you to specify manufacturer, port types, memory/memory type, chipset, etc. When I'm just starting to look for a new part or device being able to narrow down the list that way is very helpful.