Question Classification using Maximum Entropy Models

Question Answering (QA) systems are the next generation of search technology. A QA system exploits more information about a user’s intention, returning the exact answer to a users query. A crucial component of a QA system is Question Classification (QC), which involves determining the type of information the question is requesting. QC narrows the scope of processing within a QA. For example, “Who is the fastest swimmer?” the expected answer type (or classification) could be HUM:ind, which denotes a human individual. This project develops a new approach to QC, which involves the application of statistical methods, namely Maximum Entropy models for classification. This approach outperforms state-of-the-art accuracy.