Research On User Intention Change Recognition Approach Based On Situation-Theoretic

In the1970s, Belady, Lehman and others proposed the concept of software evolution, since then, most researchers have described the great deal of change of software system with software evolution. In recent years, some researchers have committed to the related research from the perspective of service evolution. As user requirement is the driving force of evolution service, so how to obtain the user requirement about service evolution quickly is the key issues that must be solved to research service evolution. Carl K. Chang and others proposed Situ framework to support service evolutioin, and they proposed a method obtaining user requirement through recognizing user intention change. This method makes user intention as the central and obtains user’s runtime requirement by detecting user intention change. Thus the process of user intention change recognition become one of the critical link in the method. Therefore, how to recognize user intention change quickly and accurately become a key problem in service evolution method drived by user intention change.To solve these problem, this thesis research problems around Smart Meeting Room System(SMR system) based on Situ framework established by our laboratory, and research user intention change recognition method in SMR system, which based on Situation-Theoretic, and supports service evolution. Firstly, in order to recognizing user intention change accurately, this thesis proposes a user intention change recognition framework based on Situation-Theoretic via researching and learning from the recent method that reasoning and recognizing user intention. After that this thesis also proposes two key problems which is action sequence pre-processing and user intention change recognition in recognition framework. Secondly, To the action sequence pre-processing problem in recognition framework, this thesis proposed a action sequence segmentation method based on scenes and maximum entropy. This method provides the input data for hidden Markov reasoning process in the Situ framework, then laids the foundation for user intention change recognition method indirectly. In addition, this method builds action knowledge tree based on scences, and builds voting rules with action knowledge tree and maximum entropy, then begins voting based on sliding window, statistics the number of votes and sets threshold, we can complete action segmentation after above processes. In addition, The optimal values of the parameters in this method is obtained by experiment. Thirdly, To the user intention change caused by the generation of new situation and situation sequence change, this thesis proposed recognition method respectively. Among them, To recognize the latter change, this thesis proposed intention change recognition method by determining digraph node reachability, which the digraph is marked based on bit vector marking scheme. Finally, a complete case is presented in this thesis to give the validity and feasibility of action sequence segmentation method based on scences and maximum entropy and user intention change recognition method in the process of recognizing intention change.