Pronunciation variation is under-studied in infant-directed speech, particularly for consonants. Regressive place assimilation involves a word-final alveolar stop taking the place of articulation of a following word-initial consonant. We investigated pronunciation variation in word-final alveolar stop consonants in storybooks read by forty-eight mothers in adult-directed or infant-directed styl...

In this paper a method for the automatic construction of a lexicon with multiple entries per word is described. The basic idea is to transform a reference word transcription by means of stochastic pronunciation rules that can be learned automatically. This approach already proved its potential (Cremelie & Martens, 1999), and is now brought to a much higher level of performance. Relative reducti...

Word pronunciation can be learned by inductive machine learning algorithms when it is represented as a classiication task: classify a letter within its local word context as mapping to its pronunciation. On the basis of generalization accuracy results from empirical studies, we argue that word pronunciation, particularly in complex spelling systems such as that of En-glish, should not be modell...

This paper presents a new approach for grapheme-to-phoneme conversion based on morphology. With this approach, a high accuracy can be obtained, although not for all words a transcription is achieved. The principle of this approach is to automatically decompose an existing pronunciation lexicon into morpheme-similar units called pseudo-morphological units. The pronunciation of the pseudo-morphol...

This paper describes development of ERJ (English Read by Japanese) database, which is designed to support CALL (Computer Assisted Language Learning) research. The DB is divided into two parts, English read by Japanese and that by Americans with the same reading sheets in both. The reading material is composed of four sections, prosodic/phonetic × sentence/word sections. 202 Japanese students we...

Speech recognition systems are conventionally broken up into phonemic acoustic models, pronouncing dictionaries in terms of the phonemic units in the acoustic model and language models in terms of lexical units from the pronouncing dictionary. Here we explore a new method for incorporating pronunciation probabilities into recognition systems by moving them from the pronouncing lexicon into the ...

In this paper, we describe a method to derive a phonetic pronunciation of a word using only an acoustic utterance of that word without a priori knowledge of the spelling of the word. In [5] and [6], we used a pronunciation model based on bigram statistics. Bi-gram statistics only constrain the left neighbor phone and results in phone sequences that are only pairwise appropriate. Here, we apply ...

This paper proposes a method to automatically extract keywords from baseball radio speech through LVCSR for highlight scene retrieval. For robust recognition, we employed acoustic and language model adaptation. In acoustic model adaptation, supervised and unsupervised adaptations were carried out using MLLR+MAP. By this two level adaptation, word accuracy was improved by 28%. In language model ...

The automatic evaluation of children’s reading performance by detecting and analyzing errors and disfluencies in speech is an important tool to build automatic reading tutors and to complement the current method of manual evaluations of overall reading ability in schools. A large amount of speech from children reading aloud plentiful in errors and disfluencies is needed to train acoustic, disfl...

Many Japanese language learners have a desire to learn how to speak natural sounding Japanese. Despite this, due to language transfer, they might find it difficult to attain this goal. The errors that occur due to language transfer can vary widely from first language to first language. In this paper, we look at what kinds of accents cause the greatest loss in naturalness in terms of pronunciati...