Design And Implementation Of The Heilongjiang Mobile Customer Service Automatic Question Answering System

With the user scale of Heilongjiang Company Limited for China Mobile Group growing, the existing business developing and the new business promoting,the service demands for most users on business consulting and problem complaints are increasing significantly, which results in greater pressure on customer service. Meanwhile, on the one hand the approach to the static set of frequently asked questions (FAQ) can't give the accurate answers to the user quickly, on the other hand the manual service requires simple and repetitive response frequently. Thereby, developing an automatic question answering system (QA) for customer service is urgently needed. In this way, we can reach an aim of sharing the pressure of customer service and improving the efficiency of customer service.In this paper, we use B/S structure and Apache server to design and implement The Heilongjiang Mobile Customer Service Automatic Question Answering System, which is based on FAQ in restricted domain. Its character is considering FAQ of the mobile communications as the source of answers, and searching FAQ to return question- answering pairs matching the user's query.This paper achieves the Chinese word segmentation by the algorithm of Double-Array Trie to shorten the response time, and combines the Part-of-speech tagging with the list of stop words and the list of interrogative words to achieve identifying the semantic information of questions and ascertaining the focus of questions. By the way of classifying cities, we index on FAQ to conserve the important information of FAQ. Based on that, we can achieve question classification retrieval, to enhance the speed of retrieval. The system uses questions analyzing and similarity computing to accomplish searching and matching questions, among them, this paper takes the improved vector space model for similarity computing to improve the system retrieval accuracy.Through functional testing and performance testing, results show that the system achieves the expected goals, meets the user's functional requirements, and under the condition that the collection of question-answering pairs includes questions the user needs, the accuracy on that the system's the first three returned results containing question-answering pairs the user needs is more than 85%, with the smooth network, response time is less than 1 second, which proves that this paper completed an Internet-based automatic question answering system with natural language interface, high accuracy and real-time, it can be used in mobile communications to ease the pressure on customer service.