Abstract
As the literature and experiments introduced in the preceding chapters showed, the performance for non-native speech recognition can be improved by considering the variations in pronunciation.
But we also saw that an explicit formulation of these variations e.g. as rules leads to a tradeoff between the number of variations considered and the dictionary or lattice size, which leads to confusions and again to a decrease in recognition accuracy. An unsolved problem about rules is also how to deal with insertions and deletions. A model is needed that handles pronunciation variations implicitly, and with sufficient context to cover insertions or deletions as well.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
Copyright information
© 2011 Springer-Verlag Berlin Heidelberg
About this chapter
Cite this chapter
Gruhn, R.E., Minker, W., Nakamura, S. (2011). Pronunciation HMMs. In: Statistical Pronunciation Modeling for Non-Native Speech Processing. Signals and Communication Technology. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-642-19586-0_7
Download citation
DOI: https://doi.org/10.1007/978-3-642-19586-0_7
Published:
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-642-19585-3
Online ISBN: 978-3-642-19586-0
eBook Packages: EngineeringEngineering (R0)