Paper: LM Studies On Filled Pauses In Spontaneous Medical Dictation

ACL ID N03-2028
Title LM Studies On Filled Pauses In Spontaneous Medical Dictation
Venue Human Language Technologies
Session Short Paper
Year 2003
Authors
  • Jochen Peters (Philips Research Laboratories Aachen, Aachen Germany)

We investigate the optimal LM treatment of abundant filled pauses (FP) in spontaneous monologues of a professional dictation task. Questions addressed here are (1) how to deal with FP in the LM history and (2) to which ex- tent can the LM distinguish between positions with high and low FP likelihood. Our results differ partly from observations reported on di- alogues. Discarding FP from all LM histories clearly improves the performance. Local per- plexities, entropies and word rankings at po- sitions following FP suggest that most FP in- dicate hesitations rather than restarts. Proper prediction of FP allows to distinguish FP from word positions by a doubled FP probability. Recognition experiments confirm the improve- ments found in our perplexity studies.