Research On The Context Understanding Model Of Chat Robot Based On Subject Information

With the explosive growth of information, people's desire of to get the right information quickly is much more stronger. To meet the demands which people can get the information faster and communicate with the machine more humanized, the quiz system comes into people's view. As a branch of of the quiz system, the chat robot develops rapidly.Through the analysis of the status of chat robot and the chat robot's understanding of the context, finding that only a few chat robot own understanding function of the specific sentence structure and most of robots do not have this function. After the analysis, the reason for this situation is that the chat robot do not own the memory function. They can not remember the previous chat content and the theme, and also can not judge whether the theme switch or not. And through the research of the theme which extracted from the dialogue, finding that the existing theme expression model can not be well applied to the real time conversation, and are all in progress based on the already existing conversation corpus.To solve this problem, this article analyse the chat conversation, and based on this, it proposes a layered label data model to store the information of the current chat theme. And on this base, this article gives the method flow of which the chat robot finish the context understanding. It also defines the computation rule of the data model, including the multiplication rule (Similarity calculation of two model) and the addition rule (Update the data model). To make the process of the chat more real, this article tried two layer context information cache to solve the theme cross-over problem. The context understanding model which we proposed make the conversation more real and the answer more precise.To verify the context understanding model, this article finished the " Owning Context Understanding Function" chat robot system on the basis of the open network chat robot Turing robot. For the user's input, the system will compare the answers of different chat robot, and show to the users intuitively.In the end, this article carried out the test of the Context Understanding Chat Robot. From analysis of the experimental results, the accuracy rate of the answer from the Context Understanding Chat Robot is higher than the chat robot, and the effect is remarkable especially for two sentence structures which are keyword elliptical and containing the pronoun. So the method which given by this article make the chat robot own the context understanding ability to a certain extent and enhances the user experience.