User Preference Profiling Based on Speech Recognition for Personalized Recommendation

In this paper, we propose a user preference profiling method with speech recognized information for personalized recommendation. Advances in speech recognition technology enable it to be widely incorporated in native application of several operating systems, and also be deployed as Web service application program interface. The accuracy of recognition is high and has shown good performance, however, it is not perfect at 100% yet. With the aim of realizing a personalized recommender system based on a user´s preference, we have designed and implemented a profiling method utilizing a user´s preferred terms. These important terms are extracted based on a user´s browsing behavior. In this study, we extend our method to refine the result of speech recognition and automatically incorporate it into a user preference term database for profiling purpose. By means of several experiments using our prototype, we show the feasibility of our proposed method.