ISCA Archive Interspeech 2013
ISCA Archive Interspeech 2013

Failure transitions for joint n-gram models and G2p conversion

Josef R. Novak, Nobuaki Minematsu, Keikichi Hirose

This work investigates two related issues in the area of WFST-based G2P conversion. The first is the impact that the approach utilized to convert a target word to an equivalent finite-state machine has on downstream decoding efficiency. The second issue considered is the impact that the approach utilized to represent the joint n-gram model via the WFST framework has on the speed and accuracy of the system. In the latter case two novel algorithms are proposed, which extend the work from [1] to enable the use of failure-transitions with joint n-gram models. All solutions presented in this work are available as part of the open-source, BSD-licensed Phonetisaurus G2P toolkit [2].

s C. Allauzen, M. Mohri, and B. Roark, “Generalized algorithms for constructing statistical language models,” in Proceedings of the 41st Annual Meeting on Association for Computational Linguistics, vol.1, 2003, 40–47. J. Novak, “Phonetisaurus g2p,” 2012. [Online]. Available: http://code.google.com/p/phonetisaurus


doi: 10.21437/Interspeech.2013-449

Cite as: Novak, J.R., Minematsu, N., Hirose, K. (2013) Failure transitions for joint n-gram models and G2p conversion. Proc. Interspeech 2013, 1821-1825, doi: 10.21437/Interspeech.2013-449

@inproceedings{novak13b_interspeech,
  author={Josef R. Novak and Nobuaki Minematsu and Keikichi Hirose},
  title={{Failure transitions for joint n-gram models and G2p conversion}},
  year=2013,
  booktitle={Proc. Interspeech 2013},
  pages={1821--1825},
  doi={10.21437/Interspeech.2013-449}
}