Spoken language understanding for conversational dialog systems

Summary form only given. Requirements for spoken language understanding for spoken dialog systems can range from the recognition of key words and phrases to a deeper understanding of the speaker's intentions. It is important to take account of the functional requirements of the overall system in order to determine the approaches and technologies that are most relevant for a particular type of dialog system. Within the research community the main focus over the past decade has been on a variety of data-driven methods in contrast to the hand-crafted approaches used predominantly in commercial systems. The main arguments for data-driven approaches are that they support more robust understanding and that they are ultimately more portable and less labour intensive. This tutorial reviews a number of different approaches to spoken language understanding within the context of conversational dialog systems, looking at issues such as coverage, robustness, context of deployment, and usability.