Panel on spoken dialog corpus composition and annotation for research

The goal of this forum is to provide researchers from various institutes with the opportunity to comment on a proposed NSF-sponsored data collection plan for a spoken dialog corpus. The corpus is to be used for research in speech recognition, spoken language understanding, dialog management, machine learning, and language generation. Currently, there exists a corpus with over 600 dialog interactions, collected from users using the Discoh system (from the IEEE SLT 2006 workshop) and the Conquest system (from ICSLP 2006) to obtain general information about conference services. These systems were created as part of a joint collaboration between CMU, ATT, Edinburgh and ICSI.