[Date Prev] [Date Next] [Thread Prev] [Thread Next] Indexes: Main | Date | Thread | Author

[ba-unrev-talk] Fwd: The Teoma Search engine


fyi...    (01)

>To: gbrain@listserv.vub.ac.be (Global Brain Discussion)
>From: Francis Heylighen <fheyligh@vub.ac.be>
>
>Yet one more little step closer to the Global Brain: Teoma implements 
>Kleinberg's HITS method, the counterpart to Google's PageRank, with pretty 
>impressive results.
>
>The following discussion comes from 
>http://www.searchenginewatch.com/searchday/02/sd0402-teoma.html
>
>
>------------------------------
>+ Teoma vs. Google, Round Two
>
>Search engine upstart Teoma has officially launched, and the media is once 
>again touting it as a Google killer. Here's a look behind the hype, and 
>the real reasons you'll want to add Teoma to your web search toolkit.
>
>Labeling Teoma a Google killer makes for a great headline, but is really 
>rather silly. Teoma is a very good search engine, but at this point it 
>poses very little threat to Google's dominance of the web search world.
>
>We'll save the face off comparison for later, after Teoma has had a chance 
>to prove itself. Meanwhile, let's look at some of the nifty things that 
>makes Teoma unique, and an excellent choice for many types of search queries.
>
>Teoma offers three kinds of results for each query. On the left of the 
>result page are "relevant web pages" that are similar to what other 
>engines produce. On the right are two other kinds of results: "Refine," a 
>list of "suggestions to narrow your search," and "Resources," which are 
>"link collections from experts and enthusiasts."
>
>Each set of results is useful, for different reasons. And all three types 
>of results are generated using proprietary technology that makes them 
>somewhat unique compared to other engines.
>
>Teoma's underlying technology is an extension of the HITS algorithm 
>developed by researchers at IBM several years ago. In a nutshell, the 
>search engine goes beyond traditional keyword and text analysis and seeks 
>out "hubs" and "authorities" related to your query terms -- a "social 
>network" of related content that forms a "community" about the topic.
>
>The cool thing about Teoma is that its community-seeking behavior is both 
>query-specific, and happens in real time. "Whenever you type in a query, 
>we're actually looking for the communities after you type the query," said 
>Paul Gardi, Teoma's Vice President of Search. "We're using a method called 
>dynamic rank, because there's a lot of information you can learn about 
>that page by its friends."
>
>Teoma's approach differs from Google's, which uses a similar, but more 
>static ranking system. It's also unlike the approach taken by Northern 
>Light and other engines that classify web pages based on pre-defined 
>categories.
>
>"We're going into the communities, finding the link structure of the 
>community using text structure as well," said Gardi.
>
>What does this mean in practice? How can this approach improve your search 
>results?
>
>First of all, by relying on the "authorities" within a community, Teoma 
>"relevant web pages" are generally quite useful, even for obscure topics. 
>Second, "Resources" are often link-rich pages -- pathfinders or 
>directories -- that are excellent starting points for further research on 
>a particular topic.
>
>But it's the "refine" results that are perhaps Teoma's most unique 
>feature. These links are automatically generated labels that "define" a 
>community for the query words you're using.
>
>So even if your initial query doesn't provide spot-on results, the 
>"refine" links allow you to "drill down" into a community, potentially 
>revealing information you can't easily find with traditional approaches to 
>information processing.
>
>"It's extremely valuable for the user to have something to refine. It's a 
>very different kind of refine because it's actually pulling you down 
>through the actual communities that exist," said Teoma's Gardi. 
>"Communities are getting stronger or weaker based on how the web is growing."
>
>This dynamic approach to surfacing content means that Teoma can discover 
>beginnings of a new community even for new or obscure pages. This makes it 
>an excellent companion or alternative to other search engines, including 
>Google, that tend to rely on lots of links pointing to pages to infer 
>authority.
>
>But Teoma is not a wholesale replacement for Google, nor is it an engine 
>you'll want to use exclusively. Teoma's index of 200 million pages is tiny 
>compared to most of the other major search engines. And the company 
>doesn't intend to compete on size, but rather on providing "authoritative" 
>results. "We're adding a little bit every day -- we're about halfway to 
>getting to where we need to be," said Gardi.
>
>Think of Teoma as a new type of hybrid between a search engine and a 
>directory, incorporating the best features of both. Like most search 
>engines, Teoma's scope is large enough to satisfy even the most obscure 
>information need, but without overwhelming you with millions of 
>near-matches or false drops. And like a good directory, Teoma structures 
>information in a way that facilitates browsing based on context and meaning.
>
>Bottom line: Teoma isn't a Google killer now, and likely never will be, 
>but it's still an excellent search engine for many types of queries. 
>Definitely worth adding to your web search toolkit.
>
>Teoma http://www.teoma.com
>
>Ask Jeeves Acquires Teoma The Search Engine Report, Oct. 2, 2001 
>http://searchenginewatch.com/sereport/01/10-ask.html Ask Jeeves has 
>purchased the Teoma search engine, which has attracted interest over 
>recent months as a potential relevancy challenger to Google.
>
>Teoma Tackles the Web SearchDay, Jun. 11, 2001 
>http://searchenginewatch.com/searchday/01/sd0611-teoma.html Teoma is a new 
>search engine born in the computer labs at Rutgers University that looks 
>like a serious contender for joining the major leagues.
>
>How Teoma Works 
>http://static.wc.teoma.com/docs/teoma/about/searchWithAuthority.html A 
>brief overview of the Teoma technology.
>
>Hypersearching the Web 
>http://www.sciam.com/1999/0699issue/0699raghavan.html An excellent 
>overview of the original HITS project and "social network" theory, and how 
>it can improve the overall quality of web search results.    (02)