Third International Conference on Spoken Language Processing (ICSLP 94)
This paper describes the processing strategy used by the Phoenix Spoken Language System, developed at Carnegie Mellon University. This system is designed to understand and respond to spontaneous speech for a specific domain. Considering the noisy and ill-formed nature of spontaneous speech and the uncertainty of the speech decoding process, robustness is a primary concern. Our goal is to respond appropriately to input, even though coverage is not complete. The natural language component of our system is oriented toward the extraction of information relevant to a task, and seeks to directly optimize the correctness of the extracted information, and therefore the system response. A flexible parser is used to map input onto semantic frames. Slots in frames represent the basic semantic entities known to the system. Frames associate sets of slots, representing specific items of information, with actions to be taken by the system. We have implemented a version of this system for the Air Travel Information Service (ATIS) task, and present evaluation results. This paper will also present results from experiments on integrating the understanding process i with the speech decoding process. We have used the semantic networks of the parser in conjunction with a bigram as a modified language model for the decoder.
Bibliographic reference. Ward, Wayne (1994): "Extracting information in spontaneous speech", In ICSLP-1994, 83-86.