Joint Processing and Discriminative Training for Letter-to-Phoneme Conversion
Published 1 June 2008
Sittichai Jiampojamarn, Colin Cherry, Grzegorz Kondrak
Citations87
Generate an AI Snapshot to get a quick, structured summary of this paper.
Study Snapshot
ObjectiveStudy objective
MethodsResearch methodology
PopulationPopulation studied
Sample sizeSample sizes
OutcomesStudy outcomes here
ResultsStudy results comes here
LimitationsResearch study limitations comes here
A concise AI-generated summary of the paper will appear here once you click Generate AI Snapshot.
TL;DR
The key idea is online discriminative training, which updates parameters according to a comparison of the current system output to the desired output, allowing the model to train all of its components together.
Abstract
We present a discriminative structureprediction model for the letter-to-phoneme task, a crucial step in text-to-speech processing. Our method encompasses three tasks that have been previously handled separately: input segmentation, phoneme prediction, and sequence modeling. The key idea is online discriminative training, which updates parameters according to a comparison of the current system output to the desired output, allowing us to train all of our components together. By folding the three steps of a pipeline approach into a unified dynamic programming framework, we are able to achieve substantial performance gains. Our results surpass the current state-of-the-art on six publicly available data sets representing four different languages. 1
Keywords
Computer Science
Machine LearningInstance-Based Learning Algorithms
4,117 Citations1991David W. Aha, Dennis Kibler +1 more
This paper describes how storage requirements can be significantly reduced with, at most, minor sacrifices in learning rate and classification accuracy and extends the nearest neighbor algorithm, which has large storage requirements.
Discriminative training methods for hidden Markov models
1,889 Citations2002Michael Collins
Experimental results on part-of-speech tagging and base noun phrase chunking are given, in both cases showing improvements over results for a maximum-entropy tagger.
Computers & Mathematics with ApplicationsParallel optimization: Theory, algorithms, and applications
830 Citations1998
Edinburgh Research Archive (University of Edinburgh)Issues in Building General Letter to Sound Rules
217 Citations1998Alan W. Black, Kevin Lenzo +1 more
A general framework for building letter to sound (LTS) rules from a word list in a language, which can be fully automatic, though a small amount of hand seeding can give better results.
Pronunciation modeling for improved spelling correction
190 Citations2001Kristina Toutanova, Robert C. Moore
The proposed method builds an explicit error model for word pronunciations by modeling pronunciation similarities between words and achieves a substantial performance improvement over the previous best performing models for spelling correction.
Applying Many-to-Many Alignments and Hidden Markov Models to Letter-to-Phoneme Conversion
186 Citations2007Sittichai Jiampojamarn, Grzegorz Kondrak +1 more
This work presents a novel technique of training with many-to-many alignments of letters and phonemes, and applies an HMM method in conjunction with a local classification model to predict a global phoneme sequence given a word.
North American Chapter of the Association for Computational LinguisticsImprovements in Phrase-Based Statistical Machine Translation
186 Citations2004Richard Zens, Hermann Ney
A highly efficient monotone search algorithm with a complexity linear in the input sentence length is described and translated results for three tasks: Verbmobil, Xerox and the Canadian Hansards are presented.
Computational LinguisticsA Multistrategy Approach to Improving Pronunciation by Analogy
115 Citations2000Yannick Marchand, R.I. Damper
This paper extends previous work on PbA in several directions, including full pattern matching between input letter string and dictionary entries, as well as including lexical stress in letter-to-phoneme conversion and extended the method to phoneme- to-letter conversion.
Language-Independent Data-Oriented Grapheme-to-Phoneme Conversion
73 Citations1997Walter Daelemans, Antal van den Bosch
An approach to grapheme-to-phoneme conversion that is both language-independent and data-oriented is described and its performance is compared to knowledge-based and alternative data- oriented approaches.
Investigations on joint-multigram models for grapheme-to-phoneme conversion
66 Citations2002M. Bisani, Hermann Ney
The joint-multigram approach to the alignment problem and use standard languagemodellingtechniquestomodellingtechniquestomodeltranscriptionprobabilities to build a grapheme-to-phoneme converter performs better than previously published ones.
Natural Language EngineeringCan syllabification improve pronunciation by analogy of English?
38 Citations2006Yannick Marchand, R.I. Damper
It is shown that including perfect syllable boundary information in the orthographic input can dramatically improve the performance of pronunciation by analogy of English words, but such information would not be available to a practical system.
Phonological Constraints and Morphological Preprocessing for Grapheme-to-Phoneme Conversion
36 Citations2007Vera Demberg, Helmut Schmid +1 more
It is shown that adding simple syllabification and stress assignment constraints, namely ‘one nucleus per syllable’ and ‘ one main stress per word’, to a joint n-gram model for g2p conversion leads to a dramatic improvement in conversion accuracy.
International Journal of Speech TechnologyAligning Text and Phonemes for Speech Technology Applications Using an EM-Like Algorithm
30 Citations2005R.I. Damper, Yannick Marchand +2 more
This paper describes the use of a form of the expectation-maximization (EM) algorithm to learn alignments of English text and phonemes, starting from a variety of initializations and finds excellent performance—the best so far reported in the literature.
Learning pronunciation dictionaries
24 Citations2006John Kominek, Alan W. Black
This paper presents an active-learning word selection strategy that is mindful of human limitations and learning rates approach that of an oracle system that knows the final LTS rule set.
A perspective on the next challenges for TTS research
21 Citations2004Juergen Schroeter, Alistair Conkie +7 more
This paper tries to identify unsolved issues in speech synthesis, the solution of which would greatly enhance the state of the art in TTS.
Rapid Development of an Afrikaans-English Speech-to-Speech Translator
18 Citations2005Herman A. Engelbrecht, Tanja Schultz
The resulting system represent the first fully functional prototype built for Afrikaans to English speech translation, and the steps required to rapidly adapt ASR, MT and TTS component to AFrikaans under limited time and data constraints.
Do not forget
16 Citations1998Antal van den Bosch, Walter Daelemans
It is concluded that keeping full memory of types rather than tokens, and excluding minority ambiguities appear to be the only performance-preserving optimisations of memory-based learning.
Improved morpho-phonological sequence processing with constraint satisfaction inference
14 Citations2006Antal van den Bosch, Sander Canisius
This work presents a global sequence-processing method that repairs inconsistent local decisions and demonstrates significant improvements in terms of word accuracy on English and Dutch letter-phoneme conversion and morphological segmentation.