Tackling new frontiers in computer science

May 27, 2004, vol. 30, no. 3
By Felicity Stone

Document Tools

Print This Article

E-mail This Page

Font Size
S      M      L      XL

Related Stories

Manuel Zahariev wants computers to understand what people say and write.

“The easy problems in computer science have been solved already,” he says. “Natural language processing is the frontier.”

Zahariev knows a few things about both language and computers. Not only is he fluent in English, French, Romanian and Bulgarian with a knowledge of Serbian and Hungarian, he recently completed a PhD in computing science from SFU.

He will receive the dean of graduate studies medal as the top student in applied sciences. He explains that although internet search engines can recognize keywords or phrases, they understand almost nothing of the text they are searching from.

He has designed software to address one specific type of information extraction: identifying what a particular acronym means from text that discusses it.

So many new acronyms are being created (think SARS, for example) that dictionaries can not keep up.

Now that this program is 92 to 95 per cent accurate, he plans to move on to matching the names of products and people to information on the web. Finding information about people is especially difficult.

Through his software company, Zahariev would like to commercialize the results of his thesis, building more knowledgeable search engines.

Search SFU News Online