Follow Slashdot stories on Twitter


Forgot your password?
Google Businesses The Internet

Google Labs Offers Table-Based Search Results 165

blackbearnh writes "Google just released Google Squared into the Google Labs playground. Google Squared lets you get results back in row and column format, and then add more columns to the result set. There's a brief tour of the features over on O'Reilly Radar, where the judgement is that there's lots of rough edges, but a huge amount of potential, especially for quick and dirty table generation for reports."
This discussion has been archived. No new comments can be posted.

Google Labs Offers Table-Based Search Results

Comments Filter:
  • Bible Books (Score:5, Funny)

    by bwalling ( 195998 ) on Thursday June 04, 2009 @09:34AM (#28208573) Homepage [] Who knew Esther was a babe?
    • Re:Bible Books (Score:4, Interesting)

      by Anonymous Coward on Thursday June 04, 2009 @09:54AM (#28208859) [] Who knew Esther was a babe?

      Uh, the thing that stuck out at me from that list was that the book of Revelations is apparently published by Ubisoft and is preceded by the book of Devastation. How did it determine that? Why, Wikipedia's list of Xbox games, of course! I don't recall that book of the Bible from the Catholic Masses I attended as a child but it sounds pretty bitching.

      If someone made games out of the books of the Bible, I'd definitely hit up Revelations (and not that Left Behind crap) but I'd assume books like Psalms and Job would be a grind :/

      • Psalms Hero!

      • by treeves ( 963993 )
        Actually, it's Revelation (or Revelation to John), not Revelations. It's a common mistake.

        Joshua would be pretty cool. Sending in spies to a prostitute's house, blasting horns and tearing down the walls of Jericho, going in to pillage, etc.

        Judges would be a pretty nasty game what with cutting up that woman who was gang-raped and sending her parts all over Israel and what not.

    • Re:Bible Books (Score:5, Informative)

      by hansamurai ( 907719 ) <> on Thursday June 04, 2009 @10:14AM (#28209115) Homepage Journal

      Well, she was a concubine of Xerxes (of 300 fame) and later his wife chosen because of her extraordinary beauty (and intelligence, but whatever).

    • by xtracto ( 837672 ) []

      The descriptions are not really useful...

    • Re:Bible Books (Score:5, Insightful)

      by treeves ( 963993 ) on Thursday June 04, 2009 @02:34PM (#28212915) Homepage Journal
      Anyone who read the book. That's who.

      Esther 2:7

      Mordecai had a cousin named Hadassah, whom he had brought up because she had neither father nor mother. This girl, who was also known as Esther, was lovely in form and features, and Mordecai had taken her as his own daughter when her father and mother died.

      Esther 2:17

      Now the king was attracted to Esther more than to any of the other women, and she won his favor and approval more than any of the other virgins. So he set a royal crown on her head and made her queen instead of Vashti.

  • How it works (Score:4, Informative)

    by Reality Master 101 ( 179095 ) <RealityMaster101 AT gmail DOT com> on Thursday June 04, 2009 @09:34AM (#28208575) Homepage Journal
    This actually seems to be a pretty simple concept. It takes the keywords on the y axis (which is the initial search), and then generates popular keywords for the x axis, and then does searches for the combinations to fill in the boxes. What goes in the box is the least amount of the target page that more-or-less fulfills the keyword search. So as near as I can tell, there's no "semantic" analysis here, it's basically a bunch of mini web searches in a grid format. It's an interesting concept, but I don't see it as any sort of world changing function, like the hype seemed to imply.
    • I'm a little confused why you would say there is no semantic analysis - perhaps it depends on the hits/searches.

      I typed in "chrome"

      And the result was:
      Item Name: Google Chrome Image: [browser screenshot] Size: A A A (WTF?) Preview Release: No value found License: Freeware

      That doesn't seem like additional search terms to me.

      • You can add "OS" as another column, then it will tell you it runs on Windows 2000 and XP.

        Try the same for Konqueror, and it tells you it runs on "No", whatever that is.

        • Heh. I tried:
          firefox, safari, internet explorer

          And it returned "Opera"

          Just firefox, safari failed oddly

          I discovered after some experimenting that adding the terms *after* the grid was created seemed to work better.
          For example first typing "firefox"
          Then google chrome

          Using their suggestion list seemed to help too.

          Anyway, my eventual grid gave, for Operating System (using their autosuggest for OS):
          Mozilla Firefox: Mac Google Chrome: [no value] Microsoft Internet Explorer: [no value] Safari: [no value] Konqu

          • I think you're missing the point somewhat. Try entering the search term "web browsers".

            Not that I'm claiming it works perfectly or anything. Apart from Firefox (developed by the Mozilla Corporation), there appears to be another web browser named Mozilla, developed by a company called "website".

        • []

          seems to have determined that they are of the family ursidae and class mammalia w/ reasonable consensus.

    • Re: (Score:2, Interesting)

      by pete-wilko ( 628329 )
      Have you tried different queries? I think the selection of the column names is actually a very difficult task and it seems to do a decent job of extracting from different pages relevant pieces of information for each column.

      If the column 'types' were known a priori then this wouldn't be that neat, however if its classifying on the fly what columns are to be used then that's pretty cool. Looks like a really nice large scale application of 'wrapper induction'.

      How the columns are determined is the impressi
      • Do you have evidence it's not just looking for the words that are common to all (or most) of the search results that come up? I'm not saying there aren't some subtleties under the hood that are pretty tricky (I don't want to be "that guy" who thinks everything could be whipped out in a couple of hours), but it seems like looking for common terms that also happen to be popular would give you a fairly good result for auto-generated keywords.
        • Do you have evidence it's not just looking for the words that are common to all (or most) of the search results that come up?

          Out of curiosity, do you work in NLP?

        • No direct evidence at all - but just looking at the different queries.

          For instance, first query I tried was "2.1 speakers" - the columns were: name, manufacturer, description, system components and speaker type (i.e. active, passive).

          Next was "normal distribution" - so not a product type search at all, and I got back: name, description, matrix-valued, degenerate, continuous. Now in fairness most of those columns were not populated, but its interesting that they were generated as they are mostly relevan
          • Re: (Score:3, Interesting)

            If you do a regular google search [] for "normal distribution", you can see the pages that come up. That it can find those keywords is not all that surprising.

            Okay, now this is interesting. Compare a regular google search of "black cat" to the Google squared one. The Google squared one pulled a whole slew of Manga results, which is not the dominant search in regular Google. That tells me that Google pulls the first X pages and tries to find pages with some sort of commonality. "Black cat" fireworks was a uni

        • by maxume ( 22995 )

          Isn't that semantic analysis? I mean, if your point is that it isn't very sophisticated, then sure, no one is likely to argue with that, but it seems to me that just using a dictionary and 'associating' a word with the words used to define it would still be semantic analysis.

        • by Zerth ( 26112 )

          It does an interesting job with just "quote" []

    • Indeed. Sounds like an application of NMF: []

  • E=MC^2 (Score:5, Interesting)

    by Celeste R ( 1002377 ) on Thursday June 04, 2009 @09:36AM (#28208593)

    It doesn't take an Einstein to find out that this is good for researching things. It certainly beats going through all of the connecting websites to get to the juicy details.

    • Too bad about its accuracy though. The data is not verified or curated like in Wolfram Alpha, so it's pretty much useless for any purpose other than leisure IMHO. It can also not do anything with the data, unlike WA. It can't even sort it. All it has going for it is that it has a lot of data. But quantity above quality for sure.

      • Re:E=MC^2 (Score:4, Interesting)

        by Anonymous Coward on Thursday June 04, 2009 @10:38AM (#28209527)

        Too bad about its accuracy though.

        Yup. I tested it out with a search on operating systems [], and there are a number of hilarious misses within the results.

        For Windows, it's apparently under a free license, and it's date of birth looks like Google scraped the drop boxes for a sign-up form rather than getting the actual creation date of Windows.

        For Linux, it was a bit off course and grabbed a description of Ubuntu, instead. It lists the current version as 2.1 which doesn't make sense for either Ubuntu (9.04) or the Linux kernel ( or 2.6.30-rc8). I suspect it grabbed the version of a random Linux app.

        Darwin resulted in the biggest miss, as one might suspect. It grabbed the biography and birthday of Charles Darwin.

    • Re: (Score:3, Interesting)

      by afxgrin ( 208686 )

      They have a long way to go then as it seems to provide some awfully terrible results. Like check this one out for Roller coasters [].

      Apparently there's a roller coaster named GhostRider, it has a length of 4,533 ft, height of 118 ft, and it travels past the speed of sound at 1038 mph!!

      I think I'll just stick to the basic Google search using quotes, +'s, -'s, AND's and OR's.

  • First thoughts (Score:5, Interesting)

    by unfasten ( 1335957 ) on Thursday June 04, 2009 @09:38AM (#28208623)
    My first thoughts after trying a few of the example searches on the main page are that it seems to be aimed a bit at Wolfram Alpha. It isn't as broad as Wolfram Alpha but it is focused on giving back data sets instead of a list of search results.
    • by leuk_he ( 194174 )

      The main problem is that wolfram alpha does not state it sources, so you cannot fact check. google is closer to the orinal search engine that it points to links.

      • Wolfram Alpha does state its sources involved in generating its answer.

        • Wolfram Alpha does state its sources involved in generating its answer.

          It rather explicitly didn't originally, and still doesn't seem to (though its removed then notice that made it clear that it didn't.) It did, and still seems to, report generic lists of all the sources that may have contributed to the internal tables consulted to generate the results, not the sources of the data actually used for the results. Originally, W|A made it clear that that was what they were doing with a disclaimer in the Source

  • And just when you thought table-based design was gone! :) I find the idea interesting, but not very practical. It seems to work well for brain storming. With just a few keywords, you get a lot of results that can shed a fresh light on what you were thinking about. Well worth a try.
    • by Celeste R ( 1002377 ) on Thursday June 04, 2009 @09:59AM (#28208931)

      Unfortunately, it's not quite there for random brainstorming. It's geared toward easy fact generation. The "human" factor is almost completely removed, image linking to entries in the table is inaccurate at best (search "Planets" and you'll find Pluto the dog), and so on.

      I can see that this is a useful tool for people like say... engineers, who need to know a material's composition and properties (facts, again), but this tool is limited by the supporting databases.

      Take, for example, the fact that I can search for a consumer product, but I can't get much more than generic information.

      Links are difficult to follow, it takes more effort than needed to go somewhere. Brainstorming is easier with the vanilla Google.

      Yes, this is a useful tool, but it doesn't compare very well to Wolfram Alpha; this is a spreadsheet data generation tool, where Alpha is an analysis tool.

    • Next thing you know, (Apple|Microsoft) will make Bing^3, which will return results in div's.
  • Epic 2014 []
    Stay tuned for the news wars next year!
  • by Lord Byron II ( 671689 ) on Thursday June 04, 2009 @09:49AM (#28208783)

    In fifteen seconds of playing around with it, I already feel like I'm able to get better data and have better control than I do with Wolfram Alpha.

    • Re: (Score:3, Interesting)

      by Jugalator ( 259273 )

      The major downside being that this can only do the equivalent of "vs" searches in Wolfram Alpha; i.e. it can't calculate with the information at all. That and the sources often being dubios at best.

    • Re: (Score:3, Informative)

      by swillden ( 191260 )

      I think it is a competitor to Alpha. I don't know that I'd say it's better, though.

      Alpha operates only on "curated" data, which means that there's a lot more that can be done with the data, since Alpha understands its structure much better. Also, Alpha can do math on it, create graphs, etc.. But with Alpha, you can only use the data that's been made available.

      Google squared can't do as much with the data, but it can use the whole web, so it's data source is much richer.

      Of course, alpha can add data

      • by Ed Avis ( 5917 ) <> on Thursday June 04, 2009 @10:26AM (#28209319) Homepage

        Well, I typed in geniuses [] and it completely failed to mention the wisest human alive, Steven Wolfram. So I think it's pretty darned incomplete, especially compared to such an unprecedented knowledge-processing breakthrough as Wolfram Alpha.

        • Re: (Score:1, Interesting)

          by Anonymous Coward

          The neat part is that, if you search for geniuses, you can add a column IQ, and get the corresponding value for this person.

          It really look promising.

        • and I typed in microsoft [] and it lists Red Hat as one of the operating systems on the right hand column under Windows.
        • This will take google bombing to the next level.
          • This will take google bombing to the next level.

            But the nice part about Squared is you can add and remove any results you want, as well as columns of information.

            You can also choose from various values for each data point, though at the moment this part seems a little limited. It would be nice to able select from more choices (maybe dynamically added as requested like other parts of the app).

    • Apples and Oranges. Google squared doesn't try to do ANY analysis of the data, it's just a way to do many searches at once in a grid format (see my other post). Alpha takes data and tries to do computations on it (with a terrible input parser, I might add).

      They aren't solving the same problem.

    • by Gulthek ( 12570 ) on Thursday June 04, 2009 @10:30AM (#28209391) Homepage Journal

      No data analysis! GS and WA are completely different beasts.

      In Google Squared:

      Try getting a square with the five largest countries by area. (In Wolfram|Alpha search for five largest countries by area [])

      Try to mathematically manipulate results like, say, dividing power usage of the united states by its population. (In Wolfram|Alpha search for united states electric production / population []

      Try to get GS to do anything like growth charts, ISS location calculations, morse code translation, puzzle solving, food calorie counting, differential equations.

      Also the data is much less complete. Check out Google Squared's results for the escape velocities of the moons of Mars []. Now check Wolfram|Alpha's []. Yeah, there's a reason that WA is citable as a primary source.

      • Also the data is much less complete. Check out Google Squared's results for the escape velocities of the moons of Mars. Now check Wolfram|Alpha's. Yeah, there's a reason that WA is citable as a primary source.

        You being a blatant shill for Wolfram notwithstanding, WA is not a primary source. Being a primary source means more than just not citing where you get your data (which I'll admit WA is excellent at). It means that you observed the data directly.

        WA is not doing original research. It is no more an origi

        • by Gulthek ( 12570 )

          Shill? Damn, if only. Just a fanboy.

          From their faq:

          Should I cite Wolfram|Alpha when I use results from it?
          Yes. For academic purposes, Wolfram|Alpha is a primary source.

          Of course, I believe we are using different definitions of "primary source []".

          A primary source could be a first-hand source from the past, such as a diary or artifact. Primary sources have been described as those sources closest to the origin of the information or idea under study.

          Me (and Wolfram|Alpha):
          Primary sources have been said to pro

  • Wolfram Alpha (Score:4, Insightful)

    by Yvanhoe ( 564877 ) on Thursday June 04, 2009 @09:52AM (#28208839) Journal
    Wolfram Alpha may not be a direct competitor for Google, yet, this is their response.
    Yes, I know, Wolfram takes info from reduced and trusted sources while Google does not. But the semantic database that they are building have the same structure.
  • by CopaceticOpus ( 965603 ) on Thursday June 04, 2009 @09:54AM (#28208855)

    Yahoo URL: "No value found" []

    Random fun: [] []

    Reply if you find other amusing queries!

  • wikipedia (Score:1, Interesting)

    by Anonymous Coward

    it sure provides a ton of information from wikipedia. i wonder what % of wikipedia articles form google's results these days.

  • Where's the option to save it to my hard drive as a spreadsheet?
  • Mixed results (Score:5, Interesting)

    by gaspyy ( 514539 ) on Thursday June 04, 2009 @10:07AM (#28209017)

    The default result sets are more than useless - are laughable.
    I searched for europe demographics [] and it automatically created a set of rows that was made of Gibraltar, Isle of Man and Faroe Islands; for columns it created Image, Description, Language, Capital and Currency. The same search on Wolfram Alpha [] produced clear, concise results.

    Eventually, I could get good results on Squared too by starting with an empty square and adding rows and columns myself. Took about 10 minutes; I could have made a simple search to get the same results.

    I realize Google-bashing is dangerous around here, but they definitely have to improve Squared if they want it to be useful.

    • Remember what Google is good at? Collective knowledge. Over time they will understand more about how people use Squared and what columns people find relevant for certain types of search, and can use that to optimize the engine. It's like the gmail spam filter which performs so well because of collective knowledge.

    • If you try "programming languages", "supreme court justices", or "linux distros" you get no results from Alpha, but a reasonable effort from Squared. Then again, these are all topics that are well covered by Wikipedia. It seems that when you do get results from Alpha, they're almost always exactly what you're looking for (unless there's a periodical with the exact phrase you're searching for, e.g. "construction equipment"). On the other hand, Squared takes more of a bad-data-is-better-than-no-data approach.
  • I get better information from a normal search. This query set is very limited (for now at least). You can only give it a broad set topic and it will only give you back more specific subsets. Nothing related, similar or tangential. Do a search for "javascript" and you get 2 results - the international standard and a link to Mozilla. Nothing about any of the popular libraries, help sites, documentation, blogs, books or history...

    Do a search on Dog breeds however and you get a nice list of those as they are a

    • SO tried again with "javascript library" and got results... but still, requiring 2 keywords to get results is hardly going improve search.

      A stop gap solution would be to suggest queries that do have a lot of results when someone is typing in the query or after a low result set is returned. Yes fewer results seems like it would be better but not for a general query... still needs work (hence why it's in Labs I suppose).

  • So I squared Microsoft and lol'd []
    License: GPL, Free ... :)

  • Spider Man = Monkey (Score:3, Interesting)

    by Celeste R ( 1002377 ) on Thursday June 04, 2009 @10:17AM (#28209167)
  • Wow. Pretty cool. (Score:5, Informative)

    by sootman ( 158191 ) on Thursday June 04, 2009 @10:22AM (#28209261) Homepage Journal

    But when I asked for [] I couldn't help but notice it was missing a few key columns.

    HOLY CRAP! This post started out as a joke but then I then typed 'measurements' into the 'Add columns' box and it effing worked! Then click in the 'Add items' in the lower left, add the 5 suggestions, do that a few more times, and BAM, you've got a good amount of data. Holy crap, this is neat. There goes the rest of my day. I could see using this for actual work, like bridge lengths and building heights and such.

    And it's FUN! Data appears instantly, as if by magic, complete with pictures. I've never said this in my 10+ years on Slashdot, but everyone, GO RTFA! Actually, skip TFA, just go visit the site!

    • by sootman ( 158191 )

      This is super cool and I'm really looking forward to what will come of it, partly because I can see this has miles of potential and partly because the results at the moment can be unintentionally comical. Searching for 'planets' yields 7: Earth, Jupiter, Pluto, Saturn, Mercury, Ceres, and Venus. And in the "description" column for Venus it says "Only at Venus, find the sexiest women's swimwear and clothing. Shop online or request a catalog for sizzling hot clothing." :-) And Pluto, evidently, has a giant si

    • Searching for supermodels? Why not search for women who actually look good instead?

  • So I search on "space physics virtual observatory" ... and it prompts me to give 5 examples so I do (VSO, VMO, VHO, ViTMO, ViRBO) ... and well, it populated a grid with the top result for "VSO" which is "VSO Software" not "The Virtual Solar Observatory" (in all fairness, the Virtual Solar Observatory doesn't show up under the search, "Space Physics Virtual Observatory VSO", but it does show up under "Virtual Observatory VSO" ... so I start with the search "Virtual Observatory" (which brings up astronomy VO

  • Not usable at all... (Score:3, Informative)

    by Arrawa ( 681474 ) on Thursday June 04, 2009 @10:37AM (#28209515)
    I've tried several searches and found that all searches are completely false, misleading or screwed up.

    Example 1: Dutch provinces. Wolfram accurately lists 12 and has the right names. Google lists dozens results, including Belgium rivers, shows the picture of a soccer player (with the same name as a

    Example 2: Dutch prime ministers. Wolfram shows the current one correctly and some older ones. All the info Wolfram shows is correct. Google lists many. Mostly the names are correct, but there is a picture and description of a car salesman with the same name, among others.

    Example 3: Countries in the EU. Wolfram shows 27 correct names. Google shows lots and lots of names. On the first pages it is ok, but on page threee, Sports is listed as an country (with the capital listed as $9500 ??) as well as Switzerland (not a EU-member) and English.

    So the tables are completely useless, it also sources Wikipedia almost all the time.

    Ergo: do not use it. Not yet in any case.
    • Re: (Score:3, Funny)

      by Virak ( 897071 )

      Obviously that's because you're searching for the wrong things. It simply can't be bothered to gather good information for such trivial matters. If you search for something worthwhile, the superiority of Google Squared quickly becomes apparent.

      Search for "list of pokemon" on Wolfram Alpha, and you get this pathetic sight []. On the other hand, if you put the same query into Google Squared, add a couple of suggested columns, and maybe a Pokemon or two you want specific information on, and it gives you something []

      • But it seems very confused about what it is: []
        • Re: (Score:3, Interesting)

          by Virak ( 897071 )

          It's Google Squared, not plain Google. If you search for something the results of which cannot be reasonably be put into a table of things and facts about them, you're not likely to get good results. A lot of people don't seem to be catching onto this. For example, in TFA, the guy searches for "atomic weights of elements", gets results which are half elements and half things like "Melting point", which have nonsense columns that are empty in most cases, and then has to add an "atomic weight" column anyway (

    • by dargaud ( 518470 )
      Strange, I find it very good. I typed "Antarctic Explorers" and then added a columned for "explored" and it tells me what those guys did explore, even adding multiple extra items. Granted, there are a few that went nowhere near Antarctica.
    • by glwtta ( 532858 )
      All your searches are for reference information and Wolfram is pretty good at this, for obvious stuff at least.

      Instead of "countries in eu", try "baltic countries": Wolfram will give you detailed (and correct) information for "baltic states", but nothing for "baltic countries", Google Squared will give you pretty decent results for both (it misses one of the capitals for the "countries" search, it comes up as the second "guess" though).

      That's the thing about W|A, you have to formulate your queries so
  • by Anonymous Coward

    I clicked the link on the front of the page for "US Presidents." presidents&suggest=1 for the lazy. Under full name for George Washington, it lists Mr. George Carver. You'd think for something 1) so common and 2) that they promote on the front of their main site should have 100% accurate and vetted information and not tell us that George Washington Carver, despite his wonderful accomplishments, was the first president of these United States.

    • Yeah, it could definitely be better. At least it displayed it in gray (everything in gray is of "Low Confidence").
  • We've been playing with returning tables for structured data in the datalab project, e.g.: [] Of course, data in tables is a well proven concept :-)
  • This Google thing will never catch on...
  • by Odonian ( 730378 ) on Thursday June 04, 2009 @11:57AM (#28210657)
    I wonder how they are ordering their table results. If I put in "star trek characters" for instance, I do indeed get a first set of ten that are all from ST.
    #1 is Spock (the Zachary Quinto version, but OK good)
    Kirk however is #6 after Riker, Troi, Picard, and Neelix.
    Neelix? c'mon google, that's a fail.
  • Unfortunately, this facility does not interoperate with w3m. All you can see is a "preparing" progress-bar thingy.

  • No semantic searches. Less accuracy than Wolfram. Lame. ;)
  • Looks like they're trying to take on Clusty [].

  • Go to: []

    Can we have the width in the unit system please? I have inches and millimeters all mixed up. It would be awesome if I can choose the unit. Measuring a car width in millimeters is not very practical from a user perspective.

  • Where is the ability to sort the records by columns? I mean duh?!?

  • If done (and used) right, this could be a quantum leap (yes, I work for Intel) in Internet search for semi-structured information search. However, my first search: 7 wonders [], gave only partially satisfactory results. Stonehenge 120 yards tall? The Empire State building and Panama channel? Well, it's rightly beta, so I shouldn't complain.

  • Kinda of like Wolfram Alpha except:
    1) less accurate.
    2) actually usable.

    Makes me think I'll be using this.

The party adjourned to a hot tub, yes. Fully clothed, I might add. -- IBM employee, testifying in California State Supreme Court