UL computer scientist Vijay Raghavan is working with collaborators at SUNY-Binghamton and Illinois-Chicago to develop an improved "metasearch engine" that is poised to compete with the larger, better-known search engines. ultoday.com talked with Dr. Raghavan about his work.

University of Louisiana, CACS, Computer Science, Vijay Raghavan

Tell us about this metasearch engine you're involved with.

We've been working on it for about 5 years. It started off as university research. When we reported our intellectual properties, we founded a company that obtained NSF funding to develop it further.

We released one of our first products in late in '07, allinonenews, but basically we're in a lack of funding mode, with all of the recent economic problems.

There is still one more product that we are close to releasing, hopefully in the next three weeks.That is MySearchView, which has drawn some attention.

What does MySearchView do?

It's a way for people who are not technologically strong to be able to create a search engine that integrates existing search engines. So you can produce results from multiple search engines.  In the very early stages, we called it "Search Engine Lego," because you can build something with basic pieces, just like kids build something from Legos.

This will produce more results, but what is better about the order in which they are reported?

It's partly coverage, but more importantly it's the depth. Sometimes there is info behind a database that is available to search engines, but not as a web page. The technology we use is basically software access to search engines. We don't do any crawling, that's why we call it a "metasearch engine." It's a type of middleware, which accepts queries from users, distributes them to different search engines, then parses and ranks the results.

A current theme in research is that "nine heads are better than one," so we are leveraging the assets in different search engines.  It's like asking many experts, and putting the answers together. We produce a superior result.

How is this different from mamma and dogpile?

It's related, but those existing technologies don't scale up. We want to be able to scale up to very large numbers, incorporating many results. For instance, with our AllInOneNews, we integrate 1200 search engines.

Anything else?

When you talk about integration & scale, it's a matter of numbers. At the webpage level, you're dealing with billions and billions of pages, crawling each one, and recording when they are updated.

By scaling up, by compiling from other search engines, we can go a level above that. There are some potential problems, because we rely on other search engines and they can pull the plug on us. But from a technology point of view, I think we can scale this up to be useful for many decades in the future.

You are working with a researcher at the University of Illinois-Chicago.

That's Clement Yu, he was my PhD advisor at the University of Alberta. He moved to UIC about the time I graduated. Weiyi Meng, our partner at at SUNY-Binghamton, is also one of his students.