[BITList] Computers figuring out what words mean

John Feltham wulguru.wantok at gmail.com
Fri Sep 19 02:42:04 BST 2008



Begin forwarded message:

From: "Gordon Barlow" <barlow at candw.ky>
Date: 19 September 2008 2:13:17 AM
To: "WORDCZAR" <WordczaRus at yahoogroups.com>
Subject: [WordczaRus] Computers figuring out what words mean
Reply-To: WordczaRus at yahoogroups.com

	
This article may interest some of you.  I see computers will be able  
to use more than a thousand words (first paragraph).  That can't be  
bad!  Gordon


The Internet got smarter this week with the release of a semantic map  
that teaches computers the meanings behind words -- and gives the  
machines a vocabulary far larger than that of a typical US college  
graduate.
Cognition Technologies began licensing the map Tuesday to software  
creators interested in having programs "understand" words based on  
tenses and sentence context -- in much the same way as the human brain  
does.

"We have taught the computer virtually all the meanings of words and  
phrases in the English language," Cognition chief executive Scott  
Jarus told AFP.

"This is clearly a building block for Web 3.0, or what is known as the  
Semantic Web. It has taken 30 years; it is a labor of love," Jarus said.

The semantic map is reportedly the world's largest, and gives  
computers a vocabulary more than 10 times as extensive as that of a  
typical US college graduate.

The coming third generation of life online is predicted to feature  
intuitive artificial intelligence applications that work swiftly  
across broadbandInternet connections.

When applied to Internet searches, semantic technology delivers  
results oriented to what people seem to be seeking instead of simply  
matching words used to online content.

For example, a semantic online search for "melancholy songs with  
birds" would know to link sadness in lyrics with various species of  
birds.

Cognition's semantic map is already used in a LexisNexis Concordance  
"e-discovery" software to sift through documents amassed during  
evidence phases of trials.

"We help them find the needle in a haystack," Jarus said.

"It used to be boxes and boxes of paper and now 80 percent of it is  
digital. Lawyers can search for a smoking gun within that discovery  
material."

Cognition's Caselaw program uses the technology to mine more than a  
half-century of US federal court decisions for legal precedents,  
according to the company.

The semantic map is also employed in a widely-used medical database.

Cognition says it has also "semantically enabled" globally popular  
online encyclopedia Wikipedia.

A Web 3.0 target is to develop artificial intelligence "agents" that  
mine mountains of information on the Internet for material that suit  
the interests of the people they serve.

"It would be a software application constantly looking for things you  
might be interested in while accurately understanding the concepts of  
what you are looking for," Jarus said.

He described it as "artificial intelligence agents working for you on  
a push basis instead of a pull basis."

Cognition has a handful of rivals, with each firm taking its own  
approach to semantic technology.

In July US software giant Microsoft bought San Francisco-based  
Powerset, a three-year-old start-up which specializes in interpreting  
the intent of people's Internet searches instead of matching specific  
words they use.

Microsoft said it plans to use Powerset technology to enhance its free  
Live Search service, which has been mired in third place behind Google  
and Yahoo in the lucrative Internet search-related advertising arena.

Powerset's semantic search merges linguistics with engineering in a  
software platform to figure out what people are seeking based on  
questions or phrases.

Standard search engines respond to individual words in the search query.

Microsoft senior vice president of search, portal and advertising  
Satya Nadella said at the time that a third of today's online searches  
don't get people the answers they seek on the first try.

"Search engines don't understand today that 'shrub' and 'tree' are  
similar concepts," Nadella wrote in a blog posting.

"We don't understand that 'cancer' sometimes refers to a disease and  
sometimes refers to a horoscope and when a query or a webpage refers  
to which."

Financial terms of the deal were not disclosed but unconfirmed reports  
were that Microsoft may have paid as much as 100 million dollars for  
Powerset.

__._,_.___

__,_._,___

ooroo

If you don't hear the knock of opportunity - build a door.

Anon.



-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.bcn.mythic-beasts.com/pipermail/bitlist/attachments/20080919/411e9ee1/attachment-0001.shtml 


More information about the BITList mailing list