Communications of the ACM - Special issue on parallelism
From text to speech: the MITalk system
From text to speech: the MITalk system
Symbolic and Neural Learning Algorithms: An Experimental Comparison
Machine Learning
C4.5: programs for machine learning
C4.5: programs for machine learning
The acquisition of stress: a data-oriented approach
Computational Linguistics - Special issue on computational phonology
IGTree: Using Trees for Compression and Classification in Lazy LearningAlgorithms
Artificial Intelligence Review - Special issue on lazy learning
Machine Learning
Machine Learning
Data-oriented methods for grapheme-to-phoneme conversion
EACL '93 Proceedings of the sixth conference on European chapter of the Association for Computational Linguistics
A general computational model for word-form recognition and production
ACL '84 Proceedings of the 10th International Conference on Computational Linguistics and 22nd annual meeting on Association for Computational Linguistics
GRAFON: a grapheme-to-phoneme conversion system for Dutch
COLING '88 Proceedings of the 12th conference on Computational linguistics - Volume 1
Representational bias in unsupervised learning of syllable structure
CONLL '05 Proceedings of the Ninth Conference on Computational Natural Language Learning
Beyond the pipeline: discrete optimization in NLP
CONLL '05 Proceedings of the Ninth Conference on Computational Natural Language Learning
Hi-index | 0.00 |
In leading morpho-phonological theories and state-of-the-art text-to-speech systems it is assumed that word pronunciation cannot be learned or performed without in-between analyses at several abstraction levels (e.g., morphological, graphemic, phonemic, syllabic, and stress levels). We challenge this assumption for the case of English word pronunciation. Using igtree, an inductive-learning decision-tree algorithms, we train and test three word-pronunciation systems in which the number of abstraction levels (implemented as sequenced modules) is reduced from five, via three, to one. The latter system, classifying letter strings directly as mapping to phonemes with stress markers, yields significantly better generalisation accuracies than the two multi-module systems. Analyses of empirical results indicate that positive utility effects of sequencing modules are outweighed by cascading errors passed on between modules.