We describe a method for learning head-transducer models of translation automatically from examples consisting of transcribed spoken utterances and reference translations of the utterances. The method proceeds by first searching for a hierarchical alignment (specifically a synchronized dependency tree) of each training example. The alignments produced are optimal with respect to a cost function that takes into account co-occurrence statistics and the recursive decomposition of the example into aligned substrings. A probabilistic head-transducer model is then constructed from the alignments. We report results of applying the method to English-to-Spanish translation in the domain of air travel information and English-to-Japanese translation in the domain of telephone operator assistance. We also report on a variation on this model-construction method in which multi-word pairings are used in the computation of the hierarchical alignments and head transducer models.
Cite as: Alshawi, H., Bangalore, S., Douglas, S. (1998) Learning phrase-based head transduction models for translation of spoken utterances. Proc. 5th International Conference on Spoken Language Processing (ICSLP 1998), paper 0293, doi: 10.21437/ICSLP.1998-578
@inproceedings{alshawi98_icslp,
author={Hiyan Alshawi and Srinivas Bangalore and Shona Douglas},
title={{Learning phrase-based head transduction models for translation of spoken utterances}},
year=1998,
booktitle={Proc. 5th International Conference on Spoken Language Processing (ICSLP 1998)},
pages={paper 0293},
doi={10.21437/ICSLP.1998-578}
}