In this work, we present a model combination approach at the word level that aims to improve the modeling of spontaneous speech variabilities on a highly spontaneous, real life medical transcription task. The technique (1) separates speech variabilities into pre-defined classes, (2) generates speech variability specific acoustic and pronunciation models and (3) properly combines these models la...

In a recent work, we proposed an acoustic data-driven grapheme-to-phoneme (G2P) conversion approach, where the probabilistic relationship between graphemes and phonemes learned through acoustic data is used along with the orthographic transcription of words to infer the phoneme sequence. In this paper, we extend our studies to under-resourced lexicon development problem. More precisely, given a...

A significant source of errors in Automatic Speech Recognition (ASR) systems is due to pronunciation variations which occur in spontaneous and conversational speech. Usually ASR systems use a finite lexicon that provides one or more pronunciations for each word. In this paper, we focus on learning a similarity function between two pronunciations. The pronunciations can be the canonical and the ...

We describe experiments in modelling the dynamics of fluent speech in which word pronunciations are modified by neighbouring context. Based on all-phone decoding of large volumes of training data, we automatically derive new word pronunciation, and context-dependent transformation rules for phone sequences. In contrast to existing techniques, the rules can be applied even to words not in the tr...

Modelling of pronunciation variability is an important part of the acoustic model of a speech recognition system. Good pronunciation models contribute to the robustness and portability of a speech recogniser. Usually pronunciation modelling is associated with the recognition lexicon which allows a direct control of HMM selection. However, in state-of-the-art systems the use of clustering techni...

Automatic Speech Recognition (ASR) can be very useful in language learning tools in order to correct mistakes in the pronunciation of foreign words by non-native speakers. Most of the systems integrating ASR proposed on the market are just rejecting or accepting whole words or whole sentences. In this paper, we propose a method to identify the pronunciation errors at the phoneme level. Indeed, ...

This paper proposes an Optimality Theory (Prince & Smolensky, 1993) [OT]-based generator of the Interlanguage [IL]1 syllabification of Korean speakers of English. Basically, I accept the ideas of 'cyclic CON-EVAL loop' and 'locally encoded finite candidate set' proposed by Hammond (1995, 1997b). However, in order to treat some features of Korean accented English such as vowel epenthesis, segmen...

Pronunciation in spontaneous Mandarin speech tends to be much more variable than in read speech. In current recognition systems, pronunciation dictionaries usually only contain one standard pronunciation for each word, so that the amount of variability that can be modelled is very limited. Most recent research work for modelling variations in spontaneous speech focuses on the lexicon level, whi...

This paper proposes the adoption of different word-level scores in the framework of automatic pronunciation scoring using learning to rank. Six types of phone-level scores are first computed and converted to word-level scores by using average-based, vowel-based, and consonant-based methods. Different score combination methods are then used to combine these word-level scores to obtain the final ...