Dr. Jim Geller - Current Research Projects

James Geller: As I mentioned previously we’re not teaching, we’re doing research, we’re doing leading edge research and different professors have different topics. I will be selfish and just talk about my own topic.

So I’m working in two areas. One of them is called Medical Informatics and the other area is Semantic Search over Web. So let me tell you a little bit about Medical Informatics. Everybody understands that nowadays there are so many medical terms that even doctors can’t keep track of them. Try this – there is a system that stores two and a half million medical terms, now you can’t imagine that anybody can really remember this, and computers to the rescue. So we have these computerized medical terminologies and there is however a big problem because somebody builds them and think about this – you have a dictionary of two and a half million terms. It’s never going to be perfect. You will always have to correct things, you will always have to add things, it’s very difficult actually to deal with this. So the research that we’ve been doing for the last 20 years is to visualize medical terminologies in the graphical format that is easy to understand and also to find errors. So my research lab has invented different techniques how we find suspicious areas in the terminology and then suggest those areas to a human expert in medicine to correct errors in there. So this is one major branch of research.

The second research project that I’ve worked on is on semantic search. Now of course you all use Google, right? So you know that you type in the search term and many times what you get back is totally off, it’s totally not what you want. So we started with the problem, very simple problem. Somebody’s looking for Michael Jackson but not for the singer but for Michael Jackson the researcher. There’s actually a researcher Michael Jackson. You go to Google you’re going to get hundreds, thousands, ten thousand hits all about the singer. So how can you eliminate this? So we’re doing research and we model a librarian. Imagine you go to a library and there’s a librarian there and you say to the librarian “Where do I find material about Michael Jackson?” If this is a good librarian she or he will ask you “Do you mean the singer or do you mean the computer science researcher?” and depending on what you say will send you at different places. So we’ve actually built a little front end for Google where our Google offers you the choices of whoever you’re looking for, the singer or the researcher.