Experiment and evaluation of crowdsourcing model for creation of voice interaction scenario

This university has published the "MMDAgent" voice interaction system toolkit as opensource software. MMDAgent enables users to develop voice interaction system easily by editing the voice interaction scenario based on Finite State Transducer (FST) format. For increase agent´s vocabulary, it is important to describe a lot of voice interaction scenario. Until now, voice interaction scenario has been described by a small organization. So, it takes a high cost to describe a lot of voice interaction scenario. For solve this problem, we construct crowdsourcing model for creation voice interaction scenario. In this study, the works of creation of voice interaction scenario is devided and distributed to a large number of people via the internet.