61.txt

来自「This complete matlab for neural network」· 文本 代码 · 共 111 行

发信人: ccipt (北方的狼), 信区: DataMining
标  题: Teoma—A New Contender Targets Google 
发信站: 南京大学小百合站 (Wed Aug 29 10:16:37 2001)


Teoma—A New Contender Targets Google 
by Richard W. Wiggins  

------------------------------------------------------------------------------
--
August 20, 2001 — In the 2 short years since Google was in beta, it has captu
red the fancy of searchers both novice and expert. Arguably, Google is the hea
vyweight champ of search engines. Now a new contender, Teoma, has strutted int
o the ring, seeking to provide the knockout blow. Teoma is currently in beta b
ut has been garnering a lot of interest in the search community. In a telephon
e interview, Paul Gardi, president and CEO of Teoma, explained to me why he ha
s such confidence. 
Teoma sports the kind of spare look and feel of early Google. When you do a se
arch, the results set is more complicated. Gardi walked me through a search fo
r the term “abuse.” The results page displays the following three choices: 


Web pages grouped by topic 
Web pages (a more-or-less conventional hit list) 
Experts’ Links 
I asked Gardi who are the experts who select the “Experts’ Links.” “Everyo
ne on the Web,” he replied. “What we do is analyze the structure of the Web.
 We look at who links to whom and we are able to identify communities—cluster
s of Web pages and sites that are connected in some important way. From that a
nalysis, we are able to identify sites that stand out as authorities in any gi
ven community.” 
Continuing the “abuse” search, Gardi pointed out that the first item on the 
Web Pages results set was the National Institute on Alcohol Abuse and Alcoholi
sm. He had me follow “Related Topic & Experts’ Links,” which brought up a s
ubset of related sites. The first “Experts’ Links” site was the Alcohol and
 Drug Abuse Institute Library at the University of Washington. Gardi noted tha
t this site is a perfect example of the concept of an authority—a site that i
tself offers a number of links to highly relevant, high-quality sites. 

Gardi contrasts Teoma’s approach with Google’s: “Unlike Google, when we exa
mine 100,000 candidate pages, we are analyzing the link structure of all 100,0
00 pages in real time. We know much more about the hubs and authorities that m
ake up the Web. We want to deliver pages that are not just relevant, but that 
are authoritative. Global linking gives credit to every link equally. We find 
the links that count, and count them.” 

Asked if the Teoma technology was similar to IBM’s work on hubs and clusters,
 Gardi said: “Yes, we are aware of their research project called Clever, but 
our implementation is different. In fact I’m not sure whether IBM ever implem
ented a working product.” 

I noted that the topical breakdown was somewhat similar in appearance to the C
ustom Search Folders of Northern Light. Gardi replied: “Our approach is actua
lly quite different. They use a taxonomy that has been developed by human edit
ors. We are building the topic list on the fly, based on the link analysis and
 the text we are able to pull out of the sites themselves. Our topical structu
re adapts itself to new topics, to new modes of expression, or even to new lan
guages, automatically, with no need for any human intervention.” 

Can Teoma scale to handle 20 million searches a day? “Absolutely. We know tha
t we can scale to a large number of users. All we need is a fat pipe to the In
ternet. We also know that we can scale to a much larger index. We believe very
 much in quality, not quantity, so we are being careful about the size to whic
h we will grow the database. But we know we can handle 1 billion URLs—or more
.” 

“Teoma can find all sorts of communities—including bad ones. Teoma is resist
ant to index spamming—something that is a huge problem for Google.” 

Gardi said work on the technology began in 1998 and the company was formed a y
ear ago. I asked what “Teoma” means. He struggled for a second to phrase a r
eply. He laughed, “It was just a code name for the project and it seems to ha
ve evolved into our company’s name. It’s a Gaelic word meaning ‘cunning.’”
 

I asked Gardi to comment on what hardware technology drives Teoma. “We don’t
 reveal any information about that.” Pressed, all he would reveal is that Sun
 computers drive the engine. 

When will Teoma leave beta? “We aren’t giving out dates. When we’ve grown t
he database to a size we’re comfortable with, we’ll call it official. One th
ing about our community-clustering technology is that it becomes more effectiv
e as the database gets larger.” 

The growing buzz over Teoma surprises and pleases Gardi. “It’s just been ama
zing how fast word is getting around.” 

Gardi is not shy about praising his search engine. “We know our technology is
 far superior to all the others out there. Maybe 1 percent of the time other e
ngines do better than we do. We’re working on fixing that 1 percent.” 

Lou Rosenfeld, noted commentator on information architecture (http://www.louis
rosenfeld.com), said: “A good idea, maybe. Although many Web-wide search engi
nes have yet to figure it out, it’s no secret that users benefit from having 
results presented in different ways. And maybe, just maybe, using different al
gorithms will help too. Teoma has clued in, and by providing results in three 
different ways, may be on to something that delights users.” 

Can search engine history repeat itself? Can a new contender with an improbabl
e name and a better way to measure the value of Web sites knock out the curren
t heavyweight champ? Teoma’s Gardi is confident they can do just that. Seriou
s Web searchers should take a look at Teoma and provide feedback to the fledgl
ing company. 
  


--
FAMILY=(F)ATHER (A)ND (M)OTHER, (I) (L)OVE (Y)OU!

※ 来源:．南京大学小百合站 http://bbs.nju.edu.cn [FROM: 202.100.5.132]