LODESTAR: a Mandarin spoken dialogue system for travel information retrieval

Abstract

This paper described the THISL spoken document retrieval system
\nfor British and North American Broadcast News. The system is
\nbased on the ABBOT large vocabulary speech recognizer and a probabilistic
\ntext retrieval system. We discuss the development of a realtime
\nBritish English Broadcast News system, and its integration
\ninto a spoken document retrieval system. Detailed evaluation is performed
\nusing a similar North American Broadcast News system, to
\ntake advantage of the TREC SDR evaluation methodology. We report
\nresults on this evaluation, with particular reference to the effect
\nof query expansion and of automatic segmentation algorithms.