We present a model of gradient phonotactics that is shown to reduce overall phoneme uncertainty in language when the phonotactic grammar modularized an unsupervised fashion create more than one sub-grammar. Our recurrent neural network (Elman 1990), which, applied two separate, randomly initialized modules corpus Japanese words, learns lexical subdivisions closely correlate with main strata for...

Speech sounds can be characterized by articulatory features. Articulatory features are typically estimated using a set of multilayer perceptrons (MLPs), i.e., a separate MLP is trained for each articulatory feature. In this paper, we investigate multitask learning (MTL) approach for joint estimation of articulatory features with and without phoneme classification as subtask. Our studies show th...

In this paper, we analyze the confusions patterns at three places in the hybrid phoneme recognition system. The confusions are analyzed at the pronunciation, the posterior probability, and the phoneme recognizer levels. The confusions show significant structure that is similar at all levels. Some confusions also correlate with human psychoacoustic experiments in white masking noise. These struc...

Analyses of errors in the natural speech of Dutch, German, and English have shown that involuntary rearrangements of phonemes (e.g., left hemisphere heft lemisphere) are more likely to occur when the two words involved in the error have the same phoneme before or after the phoneme on which the error occurred (e.g., /Ε/ in left hemisphere) [1, 2]. A study by Dell (1984) has revealed that phoneme...

This research investigated whether acoustic-phonetic knowledge of the mother tongue of a non-native speaker can be used to adapt an existing target language phoneme HMM recognizer. For this purpose three sets of phoneme HMMs were generated, one representing the target language (German), one the mother tongue of the non-native speaker (Turkish), and the third the foreign-accented pronunciation o...

The improvement of Letter-To-Phoneme (L2P) conversion that can output the phoneme strings corresponding to Out-OfVocabulary (OOV) words, especially in English language, has become one of the most important issues in Text-To-Speech (TTS) research. In this paper, we propose a Two-Stage Neural Network (NN) based approach to solve the problem of conflicting output at a phonemic level. Both Letter a...

In order to measure reaction times for the detection of a phoneme target appropriately, the length and frequency of the word preceding the target phoneme should be controlled, but this control has been lacking in earlier studies. The addition of these controls, particularly for length, made possible the demonstration that the ambiguity of the word preceding a target phoneme did not cause an inc...

In this report, we show that Out-Of-Vocabulary items (OOVs), recognized using phoneme recognition, can be reasonably reliably transcribed orthographically using Machine Learning techniques. More specifically, (i) we show baseline performance of a machine learning approach to phoneme-to-grapheme conversion when different levels of artificial noise are added (simulating phoneme recognizer errors)...