Reasearch And Implementation On Topic Modeling For Task-specific Dialogue

Dialogue system offers a natural interface for human to interact with computer.It's one of the most important applications of artificial intelligence and natural language processing.Dialogue topic modeling aims at capturing implicit topic transferring pattern from human dialogue,which makes dialogue system more personate.Dialogue topic modeling in this paper is mainly about task-specific dialogue including customer service dialogue.Research achievement of this paper can be applied to dialogue system for custom service to improve its service quality.Therefore,research in this paper is of great value.Based on related research,main work of this thesis is as follows.Firstly,due to lack of Chinese dialogue corpus,we collect and tag some real dialogue data.We analysis this data set and conclude features of task-specific dialogue.Secondly,a model for judging dialogue topic based on maximum entropy and a model for tagging dialogue topic based on conditional random fields are come up in this paper.According to features of task-specific dialogue,feature dimension reduction and dialogue history are applied to improve above models.Lastly,a nature language understanding module is implemented based on above techniques.Dialogue system with this nature language understanding module gives an outstanding performance.