Less Web searching, more Web finding.
The founders of a new U.S. start-up called SemanticV have come up with a new weapon in the war against information overload: a search engine that actually learns the meaning of words for which it’s searching.
In June 2009, Americans conducted 14 billion Internet searches; about 40% of adult Web surfers conduct at least one in a typical day. As people post more information to the Internet – digitized public records, medical information, scanned pages from old books, etc. – the search experience could become less efficient, because the act of looking up a simple word or topic on Google in the future will yield a greater number of results than it does today.
That might sound like a good thing, but Google can’t tell you which of the myriad pages it gives you is the most relevant to the topic about which you’re seeking information. In instances when you’re researching a term or a topic that’s ambiguous, the list of results will mostly be irrelevant to what you’re seeking.
For example, the word tank could refer to a piece of military equipment or a storage container for oil. Plug the word tank into the Google query box and you’ll get a wide variety of results. If you’re not sure what type of tank you want to find, you won’t be able to add any relevant tags to narrow your search. The less you know about the subject you’re looking up, the more laborious and inefficient the research process becomes as you’re forced to spend more time going through bad leads.
SemanticV’s “Stingray” engine divines the meaning of words based on how those words are actually used, as opposed to the number of times they show up in a Web page and the popularity of that page (which is what Google does).
Stingray operates based on semantics, or the scientific study of the meaning of words. The program allows you to look up different words in different bodies of text. Wikipedia is one such body that SemanticV uses in its online demonstration; the e-mail archive of Enron is another. You, the user, would add your own, like your own company’s e-mail archive or maybe all the Google results for one particular question. The results change depending on where you’re looking.
“For example, within the scope of Wikipedia, the word tank is used in a variety of ways, mostly military,” says Aaron Barnett, one of the SemanticV founders. ?gIn the Enron e.mail archives, the word tank also has a variety of meanings and usages that are particular to how energy companies manage storage and transportation.?h
The Stingray engine doesn’t just look for the word in the text; it analyzes the surrounding words the same way a human brain trying to figure out the meaning of a new verb or noun will scrutinize the word in context, or within multiple contexts. Stingray uses the information it gathers to show you synonyms, themes, and patterns.
“If you search for God in the King James Bible, Stingray sees two strong subject areas – one heavy with thou and shalt and another having to do with Christ and apostles. While reading through movie scripts however, God is understood in another sense, as an expletive,” says Barnett.
Once you see how different people use the word you’re looking up, you can cut down on your research time by disregarding the clutter, like the expletives that are complicating your search for God. The words are rendered less ambiguous, and the search experience becomes more productive, even conversational.
“By asking for better results, choosing a meaning, you communicate with Stingray,” says Barnett.
– Patrick Tucker