Introduction to the Special Issue on Spoken Language Understanding in Conversational Systems

Understanding spoken language is about extracting the meaning from speech utterances. Although there continues to be endless debates in linguistics, philosophy, psychology, and neuroscience on what constitutes the meaning of a natural language utterance (Jackendoff, 2002), for the purpose of human– computer interactive systems, ‘‘meaning’’ is regarded as a representation that can be executed by an interpreter in order to change the state of the system. In such systems, understanding spoken language involves automatic speech recognition (ASR) and spoken language understanding (SLU)—a transduction of the recognition result to an interpretable representation. There are a number of aspects to SLU that makes this a challenging task. The most challenging of them all is the issue of meaning representation. Human language expresses meaning through a variety of surface forms, for example, prosody, lexical choice, syntax. The same meaning can be expressed in many different surface forms and also the same surface form can express many different meanings. These aspects are further accentuated in conversational systems where the dialog context plays a significant role in the meaning of an utterance. Designing a representation that captures this rich expressivity in full generality is a daunting task. So, in order to build practical systems, meaning representations tend to be crafted based on the desired capabilities