Chatting Robots

The chatting robot we developing is a concept robot which talks with people in an informal manner. It consists of at least three modules, including automatic speech recognition, natural language matching and speech synthesis. In this paper, we focused on the module of natural language matching which produces response to any given sentences. Firstly, we obtained the input sentence from speech recognition module. Then the best concept sentence which may represent the meaning of input sentence is formed under certain mapping rules. Finally, based on the concept sentence, a response sentence can be produced. Each input sentence is mapped to some concept sentences according to keywords and properties of part-of-speech. For the complex characteristics of spoken Chinese, it is important to evaluate the whole concept sentence. With training samples, the weights of different part-of-speech are obtained by orthogonal optimization method. Experiment results have positively demonstrated the feasibility of the proposed approach and validation of such a chatting system.