<< Often people search to find basic facts. Three months ago we began using Squared technology to highlight answers for these types of searches in snippets. Today we’re expanding that effort so that when you’re looking for this kind of simple fact in search, we give you more accurate answers right at the top of your results, sourced from across the web:
http://3.bp.blogspot.com/_7ZYqYi4xigk/S-rI6daxXSI/AAAAAAAAGFA/DTp8c7nvQN0/s1600/closed.png
>>
http://googleblog.blogspot.com/2010/05/understanding-web-to-find-short-answers.html |
short answers is not working for me. but I really want to test this feature. |
Jérôme's "[Wolfram Alpha Squared method] works very well." – I concurr, bordering even on the truly anally-retentive (Catherine Z-J is 40.63 years old????). The only thing missing from the derivative Squared results were Zodiac signs and placement of heavenly bodies at the time of Zeta-Jones' birth and/or projected conception dates – which, come to think of it, was missing too.
TOMHTML – in your example Google is actually ladling out erronous (outdated by a month) result; while Wolfram Alpha is sort-of incomplete. Mr. Komorowski, head of the Polish Parliament, is Acting President I believe, not a formal bearer of the title yet, if ever. Correct result would read Elected "None"; Acting: Komorowski.
Is Wolfram Alpha (re)using Google's Squared Labs' technology, or a private function clone of such? |
I'm guessing say the 'clutter' is a just is the way to test interaction and then stream line accordingly. |
In the attached post, Philipp shows how an image search for "zebra" can generate a list of related animals under the heading "something different". The animals listed are:
giraffe, hippo, elephant, cheetah and leopard.
Philipp makes an analogy with Google Sets, but if you enter "zebra" into Google sets you don't get a list of animals. Instead, you get this list:
zebra, zalman, kategorie, zyxel, marque, epson, zipling, datamax, internec, brother, pentel (plus two items in non-European characters).
However Google is generating "something different", it doesn't seem to be using the same algorithm as Google Sets. |
I think the underlying technology could well be different, so I guess it's worth nothing that I intended to say that it provides "other members of a derived group" as does Google Sets too, but not necessarily the same as Google Sets. Though, if you enter "Leopard" and "Giraffe" into Google Sets, you'll also be getting Zebra, Cheetah, and Elephant in the resulting list. Entering just Giraffe works well too in terms of deriving the group "animal", though entering "Leopard" contains too much ambuigity/ overlap with the Leopard OS. Entering just "Zebra" I'm getting a lot of weird stuff – much of it starting with the letter "z" – as well as Korean results. Wonder what's happening here? |
"Zebra" is a brand name for office products.
Regarding Google Sets: when the experiment was made public, the first thing I typed in was "John", "Paul" and "George". Google Sets added "Ringo" and returned the list of the four Beatles. I was mightily impressed.
Now, like Google Web Search restults, Google Sets results have become fuzzier. For "John", "Paul" and "George" it returns a dozen or so results, most of them generally related to the Beatles but not specifically part of a set.
On the Google Sets "Discuss" page... http://www.googlelabs.com/show_details?app_key=agtnbGFiczIwLXd3d3ITCxIMTGFic0FwcE1vZGVsGIYxDA is this curiously worded "Last Update":
March 25, 2008 – Check out a patent on technology related to sets http://patft.uspto.gov/netacgi/nph-Parser?Sect1=PTO2&Sect2=HITOFF&u=%2Fnetahtml%2FPTO%2Fsearch-adv.htm&r=1&p=1&f=G&l=50&d=PTXT&S1=7,350,187.PN.&OS=pn/7,350,187&RS=PN/7,350,187
What they really mean is "Check out OUR patent...". I wonder why they feel the need to trumpet their patent here.
The patent itself gives a clue as to why Google Sets might have become fuzzier over the years. I think the first iteration simply looked for bulleted lists on HTML pages, but the patent mentions many ways of mining web pages for possibly-related terms:
- items structured as HTML ordered lists - items structured as HTML unordered lists - items structured as HTML definition lists - items that are tagged (e.g. by CSS classes, I suppose) - items separated by tabs - items separated by commas - items separated by semicolons - items stored in a table
The HTML variants are described as "related art", so they are not covered by this patent. Most of the patent is about determining the relevance of the lists. |
I'm happy I am not in the zebra selling business, I think I would only be able to sell images of zebra's through Google...;) |
As a long time Google User, I'm very annoyed by these changes. On top of which Google isn't being the paragon of reliability that it used to be. It's stopped working perfectly.... |