Designing a Question Answering System by Providing a Hybrid Method for Answer Ranking

New information technology and Internet services are facing with the growing problem of selection and management of text information. Question answering system is a solution for identifying the correct answer among a wide range of documents. This article proposed a method for ranking answers based on a hybrid method for QA systems. In the section of answers processing, different features such as statistical features, semantic features and syntactic features are extracted for training and testing of the SVM classifier. Finally, based on the answers found to a question occurring in what category and what the final score would be based on the calculated scores of features, the answers will be rated. The answers with higher priority are provided to the user as the correct ones. The mean reciprocal rank measure was used for assessing the system. This measure with the value of 0.693 indicates a good performance of the proposed method. The high percentage of system accuracy also represents the acceptable performance of the system

کلیدواژه ها:

Question Answering System ، Answer ranking ، Natural language processing ، support vector machines