摘要

Conversation systems are moving out of the research labs and out into the real world. The persistent belief that conversation would be the most natural and powerful user interface to computers. This change is important for the performance of the systems. In this paper we present an experiment named a multimedia conversation System with application in supervised learning method and ranking function. This paper firstly discusses how to use multimedia information build knowledge database. Then it introduces applying the supervised learning method and ranking function to Natural Language Understanding (NLU) and Natural Language Generation (NLG). Finally, it gives experiment approach and results. The experiment showed that this method could achieve better results in practice.