• Corpus ID: 1935750

Online discriminative training for grapheme-to-phoneme conversion

  title={Online discriminative training for grapheme-to-phoneme conversion},
  author={Sittichai Jiampojamarn and Grzegorz Kondrak},
We present an online discriminative training approach to grapheme-to-phoneme (g2p) conversion. We employ a manyto-many alignment between graphemes and phonemes, which overcomes the limitations of widely used one-to-one alignments. The discriminative structure-prediction model incorporates input segmentation, phoneme prediction, and sequence modeling in a unified dynamic programming framework. The learning model is able to capture both local context features in inputs, as well as non-local… 

Tables from this paper

A Hybrid Approach to Grapheme-Phoneme Conversion

A simple and effective approach based on a set of manually edited grapheme-phoneme mappings which drives not only the alignment of words and corresponding pronunciations, but also the segmentation of words during model training and application, respectively.

Solving the Phoneme Conflict in Grapheme-to-Phoneme Conversion Using a Two-Stage Neural Network-Based Approach

A two-stage neural network-based approach that converts the input text to phoneme sequences in the first stage and then predicts each output phoneme in the second stage using the phonemic information obtained to improve the performance of G2P conversion.

Structured Adaptive Regularization of Weight Vectors for a Robust Grapheme-to-Phoneme Conversion Model

Adaptive Regularization of Weight Vectors (AROW) is applied on g2p conversion task which is structured learning problem and achieves a 6.8% error reduction rate compared to MIRA in terms of phoneme error rate.

Hidden Markov models with context-sensitive observations for grapheme-to-phoneme conversion

The underlying concept is to model context at the observations for HMMs with discrete observations and discrete probability distributions.

Grapheme-to-phoneme conversion based on adaptive regularization of weight vectors

Adaptive Regularization of Weight Vectors (AROW) is applied to g2p conversion and achieves a 5.3% error reduction rate compared to MIRA implemented in DirecTL+ in terms of phoneme error rate while requiring only 78% the training time.

Grapheme to Phoneme Translation Using Conditional Random Fields with Re-Ranking

This work proposes a G2P system which is optimized for producing a high-quality top-k list of candidate pronunciations for an input grapheme string and uses Conditional Random Fields (CRF) to predict phonemes from graphemes and a discriminative re-ranker, which incorporates information from previous stages in the pipeline with a graphone language model to construct ahigh-quality ranked list of results.

Grapheme-to-phoneme Conversion based on Adaptive Regularization of Weight Vectors

This paper first applies Adaptive Regularization of Weight Vectors (AROW) to g2p conversion which is structured learning problem and achieves a 5.3% error reduction rate compared to MIRA implemented in DirecTL+ in terms of phoneme error rate while requiring only 78% the training time.

Comparison of Grapheme-to-Phoneme Methods on Large Pronunciation Dictionaries and LVCSR Tasks

The focus in this paper is measuring the effect of improved G2P modeling on LVCSR performance for a challenging ASR task and using n-Best pronunciation variants instead of single best is investigated briefly.

Structured soft margin confidence weighted learning for grapheme-to-phoneme conversion

The proposed method extends multiclass CW in two ways, allowing for improved robustness to overfitting: (1) regularization inspired by soft margin support vector machines, allows for margin error, and (2) update using N-best hypotheses simultaneously and interdependently.

Incorporating alignments into Conditional Random Fields for grapheme to phoneme conversion

Two approaches to integrate the alignment generation directly and efficiently into the CRF training process are proposed, one of which relies on linear segmentation as starting point and the other considers all possible alignments given certain constraints.



Joint-sequence models for grapheme-to-phoneme conversion

Joint Processing and Discriminative Training for Letter-to-Phoneme Conversion

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.

Hidden Markov models for grapheme to phoneme conversion

The paper describes the basic HMM framework and enhancements which use preprocessing, context dependent models and a syllable level stress model, and the power of the framework lies in that training of the models is performed in a single step.

Investigations on joint-multigram models for grapheme-to-phoneme conversion

A fully data-driven, language independent way of building a grapheme-to-phoneme converter using the joint-multigram approach to the alignment problem and using standard languagemodelling techniques and corpus size in detail is presented.

Language-Independent Data-Oriented Grapheme-to-Phoneme Conversion

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.

Conditional and joint models for grapheme-to-phoneme conversion

The performance of the best model, the joint n-gram model, compares favorably with the best results for English grapheme-tophoneme conversion reported in the literature, sometimes by a wide margin.

Applying Many-to-Many Alignments and Hidden Markov Models to Letter-to-Phoneme Conversion

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.

Automatic Syllabification with Structured SVMs for Letter-to-Phoneme Conversion

We present the first English syllabification system to improve the accuracy of letter-tophoneme conversion. We propose a novel discriminative approach to automatic syllabification based on structured

A multistrategy approach to improving pronunciation by analogy

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.

Improved morpho-phonological sequence processing with constraint satisfaction inference

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.