Inspecting MSN Search 345
ins0maniac writes "I compared Yahoo, Google and MSN's image search. I noticed that, MSN's search had images from only a few sites. I searched for keywords britney spears and randomly checked few pages upto page number 20 and found that the 400 images were only from 3 domains :| 5in9.com, celebritypicturesarchive.com and nabou.com. This is totally weird as it doesn't seem like a search engine, but a collection of few online galleries." There's a number of other interesting notes in the entry about the new search engine. Also, Britney.
A revenue stream.. (Score:5, Interesting)
The real task, it would seem, would be to find a way to have the engine return the proper pictures for the proper searches (so typing in Daddy's birthday doesn't result in pictures of some 50 something dude banging some barely legal chick with a party hat on.)
Stuff like that.
Re:A revenue stream.. (Score:5, Interesting)
The galleries in question probably pay for dominance
That's what I think however I didn't see results from Corbis.com [corbis.com] (BillG's stock photo company) in any results of searches that I did. And I did search for pretty generic stuff (ie: "ansel adams" who, I believe, Corbis owns the rights to)
Lack of returned hits... (Score:4, Interesting)
search for "linux" (Score:5, Interesting)
www.microsoft.com
Windows outperforms Linux: Industry case studies and test lab results provide insight into the advantages of the Microsoft®...
Slashdotted (Score:5, Interesting)
Unless the images are titled, tagged, annotated, etc., there's no good way to index them.
If I just throws a bunch of images up on a web site, there's not good technology, other than some pretty advanced facial recognition stuff, that can determine who, or what, a particular picture represents.
Change the resolution, color depth, etc. and I change the checksum for the image, so the index fails to recognize that one picture is the "same" as another, just resized, etc.
I see a lot of that on Google's image search - but can't find a way around it, either.
it looks like... (Score:5, Interesting)
this is contrary to google image search where it's not simply searching for filenames. google search seems to understand that images of britney spears need not have "britney" and "spears" in the filename.
Re:Lack of returned hits... (Score:3, Interesting)
Re:A revenue stream.. (Score:1, Interesting)
Re:Standard MS Tactics (Score:5, Interesting)
Ex.
-IE debacle, where Microsoft played catch-up to Netscape and other existing browsers after failing to neglect thier need in earlier years.
-Direct3D, which played second fiddle to OpenGL for years in usability and features till Microsoft finally began adopting parts of OpenGL's paradigm for computer graphics.
-The modern desktop GUI. A product of Apple in many respects, but later was adopted by Microsoft.
-Powerpoint, Visio and other 'Office' products. They were created by other companies, and then consumed by Microsoft.
And the list goes on and on. Today thier trying to same with hand-held media players (derived from the success of iPods), search technologies (coming from Yahoo, Google, and other succesfull search/advertisement ventures), spyware detection and many other Microsoft 'Innovations' that are soon to hit the market.
Don't Underestimate Micro$oft (Score:4, Interesting)
The current barrier to entering the market for search engines is low. The technology is relatively simple as the multitude of search-engine companies will attest.
The advantage that M$ has, over Google, is its huge R&D budget. M$ labs is the modern-day equivalent of the venerable Bell Laboratories, which is shriveling under the management of Lucent. M$ has plucked numerous professors from the computer science departments at top universities by offering incredibly high salaries.
Re:search for "linux" (Score:5, Interesting)
Guess the search engine is not so bad after all.
Bug in MSN Search Feedback (Score:3, Interesting)
Re:search for "linux" (Score:3, Interesting)
Re:Thumbnails Don't Match (Score:3, Interesting)
You know, it's a wonder nobody has started spoofing image thumbnails by returning a different image when a Googlebot comes by.
Surfer: Mmmmm... Hot, nude bored housewives...
*click*
Website: Hello.jpg!
Explicit embedded metadata being ignored (Score:3, Interesting)
Many pictures include this sort of search-rich information, either from the camera or added manually, using cataloging software. Google's Picasa 2 [picasa.com] freeware (Windows only) embeds it's key words just so. Microsoft Research's excellent freeware (Windows only) World-Wide Media eXchange [slashdot.org] tools do the same for geo-coding photos. There are numerous other tools that can do the same, leading to a significent set of internally 'tagged' material.
So, why aren't the search engines taking advantage of this? They're already loading the images and creating thumbnails, how much extra work is it to extract any additional information in the file and use that in it's indexing too, especially compared to the potentially increased accuracy?
Re:Standard MS Tactics (Score:5, Interesting)
OVERTURE (Score:2, Interesting)
Re:Standard MS Tactics (Score:2, Interesting)
Re:I searched for keywords britney spears and ... (Score:3, Interesting)
Re:it looks like... (Score:2, Interesting)
http://robertdfeinman.com/society/search_techniqu
It's part of an overall concern on the gatekeeper effect that having only a few search engines creates.
Misleading example (Score:2, Interesting)
Personally, I like the image portion of M$N Search and find it to be extremely accurate in finding relevant images (and it really is hard for me to admit that I like something that M$ has developed).
Re:Standard MS Tactics (Score:3, Interesting)
MS-DOS a product bought from another company.
Licensed to IBM partially through having the rails at IBM greased by a friend of the Gates family on IBM's board.
Other DOS products are later coming to market because all of the IBM PC's and software were shipped and built with MS-DOS. First to market equal a temporary but real "natural monopoly"
Windows is introduced, goes no where (1.1)
New Windows (2.0) goes nowhere.
Windows 3 comes out and interest grows. About this time DR-DOS starts to make in-roads with a smaller memory footprint and better tools.
Through all early versions of Windows MS-DOS compatibility is a key requirement because they hold the dominant position in that market.
Windows 3.1 comes out with a mysterious message that indicates using anything other than MS-DOS could have dire consequences (not just opinion a court found this to be anti-competitive behavior years after the fact) 3.0, 3.1, 3.11 all contain some "peer networking" to help eat away at the Novell NOS. (not non-competetive, very shrewd)
OS/2 (first big "non-dos" OS for IBM PC's since CPM) written by MS for IBM. The enterprise market shows interest.
OS/2 version 2 comes out, hailed as the future by Microsoft and IBM.
Development of OS/2 slows, friction between IBM and MS.
IBM pulls OS/2 away from MS, because it becomes apparent that MS has been dragging it's feet so that it has time to develop a competing product.
DR-DOS begins to rebuild from the "mysterious" message in 3x versions of windows. But it is too late, Windows 95 comes out almost impossible to separate it from MS-DOS. Now the GUI is king for sure.
Windows NT comes out. Runs text mode OS/2 apps because of a shared code base.
So there you go. Get the business through inside contacts (hey it's business, it happens), screw one competitor and pay the (small) price in court later to keep your momentum. Screw a partner to buy time, ideas and capital for your next generation product...PROFIT!