The Design Implementation And Key Technologies Of Q&A System Based On Books

With the rapid development of the Internet, the ways people acquire information and knowledge become more diverse, the Q & A system based on frequently asked questions (FAQ) is one of the most effective ways. Compared with the traditional search engine, FAQ Q & A system can return the answer of the question posed by users directly, which increases the efficiency of acquiring information and knowledge greatly. Meanwhile, FAQ Q&A system is a pretty active research branch of natural language processing. The build and update of FAQ, the analysis and similarity computing of questions and so on, are the research focus and cores of FAQ Q & A system.In the digital library, there are a lot of book resources. Among them, the Q&A books focus on providing professional answers of one or more areas. There are many Q&A pares in these books, which is the indispensable and important part of FAQ. And compared with other FAQ, they can provide more professional and accurate answers.Based on the Q & A books, We firstly analysis the research status of Q & A system in detail, introduce the basic frame and theories of Q & A system, the basic word segmentation methods, topic model and the question similarity computing methods. Then we design and implement a Q & A system based on books. Our contributions are as follows:(1) Analysis the needs of Q & A system based on books. Then we analyze and design the Q & A system to meet the needs, propose the system architecture design and module design, give the realization of the entire system. The system has already been running on the internet.(2) To meet the data requirements, we pick up the Q & A books from the large book resources using regular matching technology, extract the catalog information of the books. Then we filter the noise among the catalog using string matching technology. At last, we extract the Q & A pares and build the FAQ Library.(3) Focusing on the analysis and similarity computing of questions and giving some novel question retrieval methods according to the structure, topic and short text features of question data, which includes the topic analysis of question based on BTM and the similarity computing method based on the topic and focus of question. Then we design related experiments and compare the experimental results with the traditional methods, which shows the methods we raised have higher accuracy of answering question.