Slashdot Mirror
Netscape Search to be powered by Google
UM_Maverick
writes "According to this article
at news.com, Netscape is going to replace
their Excite-powered search engine with Google technology. "
Heh. I use my google Slashbox a lot. Linux based search engines
are good.
Google, find +keyword -keyword etc ... it's just yet another single-point-of-thought-oriented brute force engine, no matter how elegant its structure and interface. Except that they be within the boolean realm of the header/body/links of an address, there is no relationship between the keywords. (Each keyword representing a point of thought sought by the user, the minimum point of thought being in fact a phrase.) In other words, all the boolean energy remains only within the header.body.link realm of a single file, which is a single point address.
But the user's need is inevitably a curvature of thought problem. The user is seeking to solve a relationship between one thought and another.
Why not make a quantum step contribution: think Riemann-type curvature-of-thought search engine. Allow the user to keyword two points of thought, as before, and also define an allowable separation or curvature-of-thought between them, such that, as in the case of url searches for example, the keywords be not only in the same file address, but possibly in two different files which are related by a common directory name, domain name, etc.
This has economic implications for Mozilla which go far beyond web searches. With an half decent search engine, Mozilla could easily replace the fat and boated and proprietary adobe acrobat. This came up for me on a magazine company (history and law) which wanted to republish and gpl older issues in html-cdrom format. The necessary element was a decent search engine, like the ones used by lexus-nexus. In a Mozilla Gymnast, the user would still enter multiple points of thought (as in the prior boolean days), but would also be able to specify a relationship distance or curvature of thought as well. Curvature of thought for literature, law, science, programming, etc. would be defined by two points of thought (a variable name for example), but the relationship distance would be sentence, paragraph (subroutine), case-report/article, chapter, book, author and date, rather than url, subdomain, domain etc. Mozilla with an small integrated lexus-type search engine could improve on adobe acrobat, and terminate the proprietary pdf format.
Next time you are struggling with a complex, hurried and meaningful search in url space, as you scan down the pages and pages of return data (no matter how complex your boolean input), look inwardly and see that what you were really looking for to start with was a relationship between two thought, or a curvature of thought. See how restricted you were because you were restricted to a brute force single-point boolean mode thinking at the start.
Meaning comes from relationships between points of thought, the curvature of thought so to speak, and not from the point of thought itself.
Think elegant, Mozilla. Think curvature of thought. Hell baby, that's how you were born.
This is good. I use google for everything. Well, everything that isn't too new, like the last several months. I hope Netscape actually reindexes & updates their databases more often than google does.
"The urge to save humanity is almost always a false front for the urge to rule." --H.L. Mencken
If you don't recall, Netscape purchased Gnuhoo/Newhoo/Open Directory ( directory.mozilla.org) a while back with the intent of tying it to Mozilla. Will OpenDir be dropped come Mozilla final in favor of google, or will the two establish some sort of relationship? For instance, would dmoz abandon its search backend in favor of Google?
This has got to be a great shot in the arm for Google. Though Google is the greatest search engine, I am surprised that Netscape is going with such an unknown (to the general public) search engine.
Anyone have any guesses as to why Netscape went with Google? Other than the obvious fact that Google is so much better than everyone else? There has to be some sort of business reason for it...
I think Google ranks a site X by finding how many other sites link to site X. Their theory, IIRC, is that the more sites which link to X, the more important X is to that particular topic and the more likely that a user searching for a site on that topic wil want to see site X. The Google ranking of the sites which link to X is also taken into consideration. So if you want a higher Google ranking, you need to get linked from more sites and more highly ranked sites. According to their FAQ, "This definition seems circular, and it is."
I'm not sure if this is a troll or not, but since I've seen DAVEO post on other articles, I'll assume not. I have to concur with some of your other respondents: you really should try using the first person some time. It's surprisingly easy and fun to use, and you will automatically get more respect for your postings and fewer people will call you a troll. This is just friendly advice.
Your right to not believe: Americans United for Separation of Church and
This is not true.
Google don't build on Kleinberg's HITS technique.
Let me try to explain.
Google start with assign every page p a rank r[p].
Then assign a weight to every link w[p][q] form page p to page q.
something like this
w[p][q]=k+r[p]/(#link on p);
then recompute the rank like
r[p]=sum(w[s][p],s={every page with links to p})
repeat to convergence.
This is how google works as far as I can remember.
I read the paper before they was remove form Stanford's server.
While Kleinberg HITS algorithm goes like this.
It assign a authority a[p] and hub h[p] rank to each page.
start by make a guess for the authority ranking.
and the calculate h[p]
h[p]=sum(a[s],s={every page p have links to});
the recalculate a[p]
a[p]=sum(h[s],s={every page which have links to page p});
repeat to convergence.
I think it is easier to spam HITS than google becourse it use the hub structure.
I am currently working with a modification to HITS that will solve the problem(I think).
By the way check out my own search engine at Aeiwi
It is not a ranking engine as Google and HITS, but a automatic generated
directory.
Go to Netscape and notice their Search box at the top says NEW.
If you do a search using the Netscape option, the URL of the search engine appears as google.netscape.com. So apparently Google has already been installed, not merely announced.
Real Hackers search by randomly typing in URLs until they find what they're looking for.
-- $SIGNATURE
Google is good, but as many has noted, it lacks comprehensiveness. The gut of Google is in its application of novel indexing technique first published by Kleinberg(http://www.cs.cornell.edu/home/kleinber/ kleinber.html) In a 1997 paper(http://www.cs.cornell.edu/home/kleinber/auth .ps). Google also benefits from a agressively engineered spidering/storage etc.
But there is at least one other notable institution appying Kleinberg's technique. Take a look at NEC's inquirus: http://inquirus.nj.nec.com/
Its slow... But it demonstrates some other Klienberg's Algorithm's properties and applications. Enjoy!