Improving Reading Comprehension Using Knowledge Model

This paper describes the work on reading comprehension system, which accepts arbitrary articles as input and then generates answers according to the questions about the article. A new method to implement reading comprehension system is proposed in this paper. There are three steps in this system. First, the article will be parsed on the paragraph, sentence and phrase level. Second, the information is extracted from all sentences, and then appended to the knowledge model. Finally, the questions are answered by using knowledge model. With the experimental corpus the accuracy rate of knowledge matching is 62.5%, and accuracy rate of question answer is 64.8% with the system knowledge model.