FSC at TREC

This is the rst time that Fitchburg State College has entered the competition and the rst project that the students had worked on of this sort. We decided that we would split our time between building an infrastructure and on speci c techniques for information processing, and to use the project to help us understand the problem better for subsequent years. Because of the time we needed to research and understand the problem and when we started we had a real time constraint. We decided to do some fast prototyping and use simple information processing techniques to test the basic infrastructure. To allow for easier insertion of more complex methods we decided on a layered approach to the information processing. A three layer approach seemed to make the most sense. The rst layer would nd the document that may have the answer. The second would try to nd the sentence that answered the question. And the third would try to extract the answer from the sentence. We wanted to try a number of di erent approaches to processing information at each layer. However because of lack of time, we only got a chance to try a few. We spent much of time experimenting with the document retrieval portion. Even when you nd a method, you need time to play with the parameters with the weight multiplier and any number of tweaks. Unfortunately, we didn't get much time to do that. 2 Approach