In recent times, mobile spoken-dialog systems have become increasingly widespread. We have developed MMDAgent, a toolkit for building voice interaction systems. Users can customize MMDAgent to their liking by writing Finite State Transducer (FST) scripts. However, we discovered that beginners find it difficult to edit FST scripts manually because the number of states become larger to describe natural dialog. To resolve these problems, in this paper, we propose a method of editing voice interaction contents using composite state. The results of experiments conducted indicate that this objective was achieved.
展开▼
