Research On Interactive Question Understanding In Intelligent Customer Service Robot

With the rapid development and the universal of the Internet, more and more people prefer shopping on the Internet because it’s convenient and fast, commodity-rich and low price. At the same time, e-shop managers have to provide a lot customer service staff to answer the questions of customers when they are shopping. If a intelligence customer service robot could answer some simple and repeated questions, it would decrease repetitive works of service staff and enhance their work efficiency. So that it could promote customer satisfaction and cut down the cost of operating a online-shop.The customer service robot is in fact an interactive question answering system, that is to say, the robot retrieves answers from knowledge base according to the questions of customers. And also, the robot could act as an advisor for service staff. To answer the questions correctly, the robot has to understand the customer intents. While the questions in the field of customer service are characterized by colloquial forms, incomplete and context related, existing techniques could not satisfy the intelligence customer service robot.To solve the problem of understanding the questions correctly in the environment of online shopping, we proposed a method for interactive question understanding method in this paper relying on the whole context of the question in a dialogue and the role of the robot. Firstly, we proposed a speech act model for dialogues of customer services in online shopping and the context of questions are analyzed based on the dialogue structures. Second, by combining topic analyzing and semantic information extraction, we proposed the multi-dimension method for question understanding in order to get the true answers of the question.Main works are as follows:(1) analyzed the corpus of dialogues in customer service; we summarized the characteristics of interactive sentences and the dialogue.(2) based on the analyzing the dialogues of information consulting in online shopping, we proposed a speech act model for dialogue structure analyzing; proposed a topic model for question understanding.(3) Combining speech act model and topic model, we proposed a question understanding method in a dialogue oriented to information consulting in online shopping.(4) A customer service robot based on this model is constructed. At last, experiments show that our method of question understanding is better than bag-of-words based method and our method promote the performance of customer service robot in an interactive environment.