Google.com 'De-mystification'

by **vivek** » Wed May 07, 2003 2:11 pm

This board might not get a reponse as those asking if its ok to ask about virginity. But seriously, I want to get some serious insight on google.com\'s indexing fundas. You see, Google does not care a damn for what you put in meta tags, neither does it look at your key words. All it does is take snap shot of whole page and then rank. I have read about the ranking procedure but it went 2 Ft above my head. On the other hand, msn search and to an extent even the new yahoo search take into account the \'description\' that you would give in the content tag (I think). No search engine seems to care about key words, except overture. I have spoken to a zillion webmasters and all have different fundas about google\'s indexing procedure. One thing I am sure, the popular belief that Google takes into account the number of webpages linking to yours is false. I have some new fundas too. What do you think?

by **SergeyBrinsCousin** » Wed May 07, 2003 2:58 pm

Who cares what technology they use and whether its sophisticated or not - the point is as of now Google serves the purpose of millions of individuals who search based on whether the words appear on the page and browse thru the first 10 or 50 links given. Google is for the masses. If you need specific stuff to research on try NorthLight. But you have to pay - most of the time to see the results - which are almost always pretty good. The great thing about Google is that it earns money - even though its just a search engine. And Brin and team have a lot of things going on behind the search engine. They develop some really nice products (like a sports shoe which converts into a roller skate with a leg jerk!) and keep their investors happy.

by **Z C is Cantonese** » Wed May 07, 2003 7:16 pm

Dont know ant Indexing but know a bit abt Roller shoes. To my knowledge, they were first introduced in Hong Kong. THey were an instant hit among kids, ofcourse !!! next day, the doctors were concerned abt the ankle bone biomechanics. Instantly, parent factor came into picture, ofcourse !! Thats what happended when Mr. Brin put his finger into what i call the \"true human technology\".

by **pal** » Wed May 07, 2003 8:47 pm

What\'s the use of indexing anyway? (Sorry for my ignorance) Who cares which page is how famous, as long as you find what u search for?!!

by **vivek** » Thu May 08, 2003 11:04 am

never mind, its because of proper indexing you are able to find the exact page containing the information you are looking for. But all this depends on the way the webmaster \'positions\' the webpages. There are a whole bunch of tags in a web page code that you are not able to see. These are vital in the sense the search engine robots look for these. MSN, overture and others make it apparent what they are looking for in a webpage code, its the google thats tricky

by **arrow** » Sun Jun 22, 2003 9:36 pm

Hi Vivek...I\'m just curious why said \"the popular belief that Google takes into account the number of webpages linking to yours is false.\". While I\'m not into web technology by any means, I know pagerank uses Markov Models for arriving at the rank of a page. Markov chains is a well established theory in Machine learning & Mathematics. In layman terms what it says is \"future depends only on present and present depends only on immediate past and so on and so forth\". In that sense page rank depends not only the number of pages linking to yours, but also depends on the pageranks of those pages linking to yours (and the pageranks of the links pointing to the pages linking to yours and so on). While this doesn\'t definitely answer all questions (the subject is abstract) I hope it helps in giveing some clues. Regards. 
ps: I\'ve to say I\'m impressed by your eye for details. I never once thought of how pagerank really worked but only came to understand it because of my acquaintance with Markove models. I use them in my work for another purpose.

by **PGW** » Mon Jun 23, 2003 8:06 am

vivek, There are a lot of PageRank Algorithms and one of them is Coordinate Matching. Coordinate matching takes words in a document as an N-dimensional vector and converts the query too as a N-dimensional vector and finds the Cosine measure between the Query terms and the Document terms. There are two ways to find measure Coordinate matching one is the Cosine measure and the other is the Okapi BM 25. (If you want to know the actual formulae then please ask me for the same). While no one is pretty sure what algorithm Google uses they have a pretty good ranking algorithm and they are constantly working on it. You might find the following URLs useful (though they do not deal with PageRank.) http://www.computer.org/micro/mi2003/m2022.pdf http://www.wired.com/news/infostructure ... 97,00.html

by **vivek** » Mon Jun 23, 2003 12:08 pm

I came across an amazing webpage explaining the Ins and Outs of google page ranks with a disclaimer that all this can be false!!. Anyway, i am not sure about those math fundas but if you guys have any website or a page then try this.Take a popular website\\\'s home page, save it in your local drive. The change all the links in that home page ( say all the 127 links in the rediff.com home page) and let\\\'em point to your homepage. Put this page in a different location and not your host server. Repeat this process for say 5 other popular website homepages and put them in different locations and mind you, do not change title,metas and keywords of the original page. Now, this should make the google bot think that a website is liking 127 links to your homepage from location A, then say a 100 more from Location B and so on. If google takes into account the number of pages linking to your webpage as a prime ranking parameter then its childs play to in be in page one. This does not happen. The raking as stated in google.com is based on the numbe of times the webpage has been visited on a given string. Period. This is what i feel. An i feel this is the reason why you are able to find the exact info for the seach string. If results were a fucntion of what we place in metas or keywords then its easy to cheat the search engine. This does not happen with google. Inviting quality responses...By the way how many of you know about the so called \\

by **arrow** » Mon Jun 23, 2003 1:43 pm

Well...what you suggested doesn\'t work, mostly because all it means you have 5 new pages (not 5x100 links as you suggested) pointing to your home page. In addition the page rank of these 5 new pages is next to nothing (but for an initial value) that you have placed on the 5 servers. The probability (or the rank) of a page appearing to be popular goes up recursively with the popularity of the pages linking to it (and the ranks of pages linking to those pages and so on). It\'s probably more easier to visualize if one understands markov chains... 
Google does say that they combine it with string matching algorithms to decide the order of a page (wouldn\'t do to depend only on rank). That\'s also why new pages will have low ranks..Hope this makes some sense... 

Interesting discussion, btw..Regards

by **Arch** » Mon Jun 23, 2003 2:31 pm

The whole board went 12\' above MY head. coz\' it aint\' my subject matter. went thru each one of the postings. It is too technical for my understanding! have a nice time discussing, u all.

by **vivek** » Mon Jun 23, 2003 3:11 pm

Will somebody explain the whole stuff on \'Markov chains\'. Cause i am looking from a practical point of view.Allround the internet, we find something called the webring concept. Its about \' I link to you and you link me\'. These webrings are so much in number that i would say almost all the grade 2 type amateur webpages are in some ring or the other. For ex. The Mp3 websites and porno websites. These guys cannot suvive without exchanging links. Now the funda is, if google lists pages based on number of inward links ( this is not the only criteria), then atleast theoritically the pages in webrings should be highly ranked. Which is clealy not the case. My point is to stress on this mis-conception.Secondly, for a given search string you might have the first page first result simply because that page contains those words in string in the exact arrangement including the cases. Sometimes you might find that the words in search string are not even present in the title of page, still the page might be the first result. My observations suggest that, even if this page is relatively new (new by google terms is less that 45 days) it is possible to get the page as a first result we find it as first result for the \'exact search string\'. So we have a contradiction. If (going by the earlier statement) new pages get lower rank then how come it comes in first page. That means, page rank and placement of webpages in results are not really connected. Then we come back to the first question, what is a page rank and what purpose does it serve?.

by **arrow** » Mon Jun 23, 2003 4:34 pm

pagerank is not a misconception...it definitely plays a part in ordering a page in search results...For example I typed in fullhyderabad in google search and I get 5th result as http://www.excite.co.uk and 75th result as http://www.totalsportsearch.com despite both the pages listing fullhyderabad.com exactly ONCE on their pages...so what makes google choose excite over totalsportsearch?? That\'s where the inbound links stuff comes in (surely excite is popular than a noname?)...same applies for search words \"cricket\" or our forum favourite \"virginity\"...invariably you will get khel.com for cricket in top of the lists even though khel.com page doesn\'t do anything special on the webpage....and for virginity you will get (well I digress...) 
 ...And \"you link me I link you\" stuff does not work if nobody else links to you and me...but if a whole lot of yous and whole lot of I\'s link to each and every you and I, then yes... page rank will assign a better rank to you and I... I\'m not saying google doesn\'t search through the contents, but once it searches through it\'s pages it ranks them according to pagerank. If you put up a page tomorrow with \"vivekthemaverick\" in there, google might find it and rank it as 1st result (just because there are very few direct references to it) but if you put up a page with \"cricket\" (or for that matter \"virginity\") I\'m sure it will take hell of a long time before your page appears before khel.com ...Regards
 

an easy read on markov chains is here 

www.math.louisville.edu/~t0ried01/chapter9.pdf

a good explanation of page rank is at 

http://www.mathworks.com/company/newsletter/clevescorner/oct02_cleve.shtml

by **vivek** » Mon Jun 23, 2003 4:50 pm

I could not agree with you more. You are right. I guess this is where search engine optimisation comes in. In the sense, SEO is a kind of work around. The funda is simple, a guy searching for \'cricket\' is not a serious info seeker. Serious guys would search for \'Indo-pak matches\' and probably is looking for a chronological event list of matches between indiaVspak. Actually my core issue was to de-mystify google.com\'s ranking methodology. I shall write more after a little reading. Till then lets keep this talk going.

Who is online