Abstract
Grapheme to Phoneme (G2P) translation is a critical step in many natural language tasks such as text-to-speech production and automatic speech recognition. Most approaches to the G2P problem ignore phonotactical constraints and syllable structure information, and they rely on simple letter window features to produce pronunciations of words. We present a G2P translator which incorporates syllable structure into the prediction pipeline during structured prediction and re-ranking. In addition, most dictionaries contain only word-to-pronunciation pairs, which is a problem when trying to use these dictionaries as training data in a structured prediction approach to G2P translation. We present a number of improvements to the process of producing high-quality alignments of these pairs for training data. Together these two contributions improve the G2P word error rate (WER) on the CMUDict dataset by ~8%, achieving a new state-of-the-art accuracy level among open-source solutions.
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
Similar content being viewed by others
References
Ash, S., Lin, D.: Grapheme to phoneme translation using conditional random fields with re-ranking. In: Sojka, P., Horák, A., Kopeček, I., Pala, K. (eds.) TSD 2016. LNCS, vol. 9924, pp. 314–325. Springer, Cham (2016). doi:10.1007/978-3-319-45510-5_36
Baayen, R.H., Piepenbrock, R., Gulikers, L.: Celex2. Linguistic Data Consortium, Philadelphia (1996)
Bisani, M., Ney, H.: Joint-sequence models for grapheme-to-phoneme conversion. Speech Commun. 50(5), 434–451 (2008)
Deligne, S., Yvon, F., Bimbot, F.: Variable-length sequence matching for phonetic transcription using joint multigrams. In: 4th European Conference on Speech Communication and Technology (1995)
Demberg, V., Schmid, H., Möhler, G.: Phonological constraints and morphological preprocessing for grapheme-to-phoneme conversion. In: Association for Computational Linguistics, vol. 45, p. 96 (2007)
Eger, S.: Do we need bigram alignment models? On the effect of alignment quality on transduction accuracy in G2P. In: Proceedings of EMNLP, vol. 18, pp. 127–136 (2015)
Greg Fast. Lingua-en-syllable (1999). http://search.cpan.org/gregfast/Lingua-EN-Syllable-0.251
Kheang, S., Katsurada, K., Iribe, Y., Nitta, T.: Solving the phoneme conflict in grapheme-to-phoneme conversion using a two-stage neural network-based approach. IEICE Trans. Inform. Syst. 97(4), 901–910 (2014)
Dennis, H.: Klatt. Review of the arpa speech understanding project. J. Acoust. Soc. Am. 62(6), 1345–1366 (1977)
Kubo, K., Kawanami, H., Saruwatari, H., Shikano, K.: Unconstrained many-to-many alignment for automatic pronunciation annotation. In: Proceedings of the APSIPA, pp. 1–4 (2011)
Lafferty, J., McCallum, A., Pereira, F.: Conditional random fields: Probabilistic models for segmenting and labeling sequence data, pp. 282–289 (2001)
McCallum, A.K.: Mallet: A machine learning for language toolkit (2002)
Novak, J.R., Minematsu, N., Hirose, K.: WFST-based grapheme-to-phoneme conversion: open source tools for alignment, model-building and decoding. In: 10th International Workshop on Finite State Methods and Natural Language Processing, p. 45 (2012)
Novak, J.R., Minematsu, N., Hirose, K., Hori, C., Kashioka, H., Dixon, P.R.: Improving WFST-based G2P conversion with alignment constraints and RNNLM n-best rescoring. In: Interspeech (2012)
Rao, K., Peng, F., Sak, H., Beaufays, F.: Grapheme-to-phoneme conversion using long short-term memory recurrent neural networks. In: IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP) (2015)
Wang, D., King, S.: Letter-to-sound pronunciation prediction using conditional random fields. IEEE Sig. Process. Lett. 18(2), 122–125 (2011)
Wang, X., Sim, K.C.: Integrating conditional random fields and joint multi-gram model with syllabic features for grapheme-to-phone conversion. In: Interspeech, pp. 2321–2325 (2013)
Weide, R.: The CMU pronunciation dictionary, release 0.7a (2014)
Wu, K., Allauzen, C., Hall, K., Riley, M., Roark, B.: Encoding linear models as weighted finite-state transducers. In: Interspeech (2014)
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2017 Springer International Publishing AG
About this paper
Cite this paper
Ash, S., Lin, D. (2017). Incorporating Syllable Phonotactics to Improve Grapheme to Phoneme Translation. In: Quesada, J., Martín Mateos , FJ., López Soto, T. (eds) Future and Emerging Trends in Language Technology. Machine Learning and Big Data. FETLT 2016. Lecture Notes in Computer Science(), vol 10341. Springer, Cham. https://doi.org/10.1007/978-3-319-69365-1_11
Download citation
DOI: https://doi.org/10.1007/978-3-319-69365-1_11
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-319-69364-4
Online ISBN: 978-3-319-69365-1
eBook Packages: Computer ScienceComputer Science (R0)