A hybrid approach for question classification in Persian automatic question answering systems

Question classification plays a major role in automatic question answering systems. The performance of a question answering system depends directly to the performance of its question classification section. A question classifier associates a label or category to each question which represents semantic class of its answer. There exist different approaches such as rule-based, machine learning and hybrid approaches for solving this problem. In this paper we have introduced a novel hybrid question classification approach for Persian closed-domain question answering systems. The proposed approach is used practically in an online automatic question answering system. The experimental results show the usefulness of combining rule-based and machine learning question classification approaches for highly inflectional languages such as Persian. We got the satisfactory results according to high number of question classes.