Dialogue Model Study And Development In Chat Bot System

As the rapid growth of the smart phones, users of mobile phone are looking for more convenient information acquisition method and better interaction experience between humans and computers. Under such circumstance, the development of chat bot system and question answering is stepping into a new era. The core of this paper is to build a sophisticated dialogue model which can assimilate knowledge easily and can use both rule based strategy and statistics based strategy to process questions. The ultimate goal of those work is to build a chat bot system which can satisfy users.The significance of dialogue model for chat bot system had been introduced in the field of question answering. Secondly, problems in current researches have been pointed out. In addition, the relevant knowledge applied to this research also has been described, which mainly includes Foundations of Statistical Natural Language Processing, Information Extraction and Leaning to rank.In the second part of this paper, we had defined the architecture of the chat bot system and introduced the related components of chat bot. Based on that, the typical process of semantics representation had been explored, including question classification, syntax analysis, named entity recognition and so on.According to the researches above, a novel dialogue model which is adept at knowledge management, knowledge application and question interpretation has been presented. Compared with existing dialogue model, the proposed model can be extendible and give out precise answers. To complete the proposed dialogue model, we had studied the ranking strategy of answer documents. Listwise kind leaning to rank approach had been ported to rank the answers of chat bot systems by us. Moreover, experiments had been run to evaluate the performance of the ported method.At last, a chat bot prototype system has been constructed. We had release the chat bot prototype system to the public who can access the service by an android app. From the analysis to the data collected, result shows that the proposed dialogue model is feasible and valid.