Dealing with unexpected words in automatic recognition of speech

Authors:
Hynek Hermansky
Affiliations:
Center for Language and Speech Processing, The Johns Hopkins University, Baltimore, Maryland and Brno University of Technology, Czech Republic
Venue:
TSD'11 Proceedings of the 14th international conference on Text, speech and dialogue
Year:
2011

Citing 4
Cited 0

Links between Markov models and multilayer perceptrons

Advances in neural information processing systems 1
Statistical methods for speech recognition

Statistical methods for speech recognition
Error-responsive feedback mechanisms for speech recognizers

Error-responsive feedback mechanisms for speech recognizers
Recovery of rare words in lecture speech

TSD'10 Proceedings of the 13th international conference on Text, speech and dialogue

Quantified Score

Hi-index	0.00

Visualization

Abstract

Unexpected words attract listener's attention. They are informationrich and getting them right is important for human communication. In the automatic recognition of speech (ASR), words that are not in the expected lexicon of the machine are typically substituted by some acoustically similar but nevertheless wrong words. The article discusses reasons for this undesirable behavior of the machine, describes some known examples of dealing with the unexpected words in human speech perception and their implications, and proposes an alternative architecture of ASR that could alleviate some of the problems with the unexpected acoustic inputs. Some published experimental results from using this alternative architecture are given.