 |
|
| CURRENT SECTION :: News |
The No. 1 Resource for UK IT Contractors: Comprehensive guides - Daily news IT contract jobs - Market rates - Forums - IT contractor network - Calculators |
|
 |
A new breed of computer with human-like intelligence is being made possible thanks to Wikipedia, the online encyclopedia, researchers say. Experts at the Technion Faculty of Computer Science say their invention could help businesses and consumers alike, with the possibility of smarter spam filters and online searches. Although the exact method behind the breakthrough is part of a recently filed patent, the process involves downloading the entire contents of Wikipedia in order to boost a PC’s intelligence. Once the Israeli-based researchers imported the content, they used machine learning techniques to teach the computer to augment texts with knowledge from Wikipedia. Related articles enrich the texts and allow the computer to draw conclusions that would not be possible without the imported information, the researchers claim. They added that their method “enables automatic use of the encyclopedia in order to significantly improve computers’ ability to deal with tasks that normally require human level intelligence.” Cited examples where improvements are tipped include automatic filtering of e-mail, routing of intelligence reports, automatic cataloging of documents and Internet searches. The success for the researchers reinforces the idea of using large online knowledge repositories, such as Wikipedia, to significantly enhance a computer’s performance. “Existing software systems for analysing documents are based on statistical analysis of word frequencies. They do not have any other knowledge of the world,” said Dr Shaul Markovitch, AI expert at the faculty. “The new method enables a computer to understand, for example, that an e-mail message containing the term ‘C-4’ is about explosives, and therefore should be brought to the attention of an intelligence officer. “As another example considering filtering junk e-mail that attempts to promote vitamin sales. Suppose the message does not contain the word ‘vitamin’ at all, but does include the word ‘riboflavin.’ “Existing software is not capable of identifying this as commercial e-mail because it does not recognise the word ‘riboflavin.’ Our system uses Wikipedia in order to understand that Riboflavin is a kind of vitamin, and is therefore capable to better understand the essence of the message.” In the experiments, the accuracy of computerized text processing was increased by tens of percentage points, Dr Markovitch said. The research is due to be presented this week in India at the International Joint Conference on Artificial Intelligence. Jan 8, 2007 Email this article Printer friendly page Previous Page
|
|
|
||||||||||||||||||||||||||||
| All content © Contractor UK Limited | [Register for News Letter] | [Privacy Statement] | [Terms of Use] | [Top of Page] |