The Technion researchers harnessed Wikipedia to increase the computer's understanding

They developed a method that allows automatic use of the encyclopedia, to substantially improve the computer's ability to handle tasks, such as automatic filtering of e-mail or intelligence filtering of documents, channeling news, automatic cataloging of documents and searching the Internet

The researchers of the Computer Science Faculty at the Technion harnessed the computerized encyclopedia Wikipedia to increase the understanding of the computer. They developed a method that enables automatic use of the encyclopedia, to substantially improve the computer's ability to handle tasks, such as automatic filtering of e-mail or intelligence filtering of documents, channeling news, automatic cataloging of documents and Internet searching.

"Without knowledge of the world, it is difficult to perform intelligent tasks on a computer," explains Dr. Shaul Markowitz, an expert in artificial intelligence and learning systems at the Faculty of Computer Science at the Technion. "And knowledge is in encyclopedias". The method was developed over four years together with doctoral student Yevgeny Gavrilovich, has already been published in two scientific conferences and the Technion registered a patent for it. The new achievements of the research will be presented at the world conference on artificial intelligence that will be held in India this coming January.

The Technion researchers brought Wikipedia to their computers, and used automatic learning techniques to allow the computer to link texts and articles in Wikipedia. The related articles enrich the texts and allow the computer to draw conclusions that would not have been possible without the additional knowledge.
"The existing programs for analyzing documents rely on statistical analysis of word occurrences. They don't have any additional knowledge about the world," explains Dr. Markowitz. "The new method allows the computer to understand, for example, that an e-mail message containing the C-4 symbol is related to explosives and therefore deserves intelligence attention. Another example is a spam filter that tries to filter messages related to the sale of vitamins. Suppose a message arrives that does not include the word vitamin at all, but does include the word 'riboflavin'. Existing programs are unable to detect such a spam message because they have no knowledge of the word 'riboflavin'. Our system will use Wikipedia to conclude that 'riboflavin' is a type of vitamin and is therefore a spam message to be filtered. Access to huge online databases, such as Wikipedia, improves computer performance considerably. In the experiments we conducted, the accuracy of the computer increased by dozens of percent."

To the Hebrew Wikipedia site