Research Challenges for the Development of Interactive Spoken Language Systems

The last five years have witnessed unprecedented progress in human language technology (HLT) － speech recognition and language understanding capabilities are improving at a very rapid rate. To meet the challenges of developing a language-based interface to help users solve real problems, however, we 1nust continue to improve the core technologies while expanding the scope of the underlying HLT base. This paper outlines my view on what are some of the unn1et research challenges, including the need to work in real do1nains, dialogue n1odelling, dealing with unknown words, spoken language generation, and portability across domains and languages.