SLU in Commercial and Research Spoken Dialogue Systems
David Suendermann, Roberto Pieraccini · 2011
Spoken language understanding (SLU) plays a key role in every spoken dialogue system. This chapter shows how different SLU techniques are integrated into dialogue systems, focusing on the significant differences between research and commercial implementations. After discussing about how to make this integration robust against speech recognition glitches, the chapter reviews example projects, architectures, and corpora associated with the application of SLU to spoken dialogue systems. The chapter provides an overview about both research and commercial aspects of SLU in dialogue systems. It reviews approaches, deals with deficiencies regarding speech input, discusses about example datasets and applications, shows how to measure performance, and finally draws the conclusion that research and commercial worlds, despite of major differences in their approaches to SLU and dialogue management, are slowly growing together. Controlled Vocabulary Terms interactive systems; learning (artificial intelligence); speech processing