Carnegie Mellon University

DialPort: Enabling Spoken Dialog Research with Real Data

Speech Processing

Technology once used exclusively in research labs with experimental subjects is now available for virtually everyone. This development puts pressure on the academic research community to both provide skilled practitioners and to tackle the next set of basic research problems that will allow the spoken dialog systems field to progress. The research community must also transition from performing relatively small-scale experiments with paid experimental subjects to embarking on more large-scale interaction with real users. We propose a transformational infrastructure that will enable more researchers to participate in this new paradigm. Specifically, we will create a Dialog Portal (DialPort) and conduct challenges that spark the creation of novel, real spoken dialog applications and bring new system creators into the community. Through this effort, we will educate students about our field, involving them in research and using their involvement to attract more students (especially minorities and women) to STEM careers; provide potential system creators with the right tools and teach them how to implement systems; create infrastructure that helps system creators access real users; and help system creators and research teams make data and software available to the community. We will develop both user and developer communities and provide resources that lower the barriers to entry for each group. Our results will spark research in areas such as virtual agents, robotics and communication, human-computer interaction, among others.