In this work, new multi-classifier schemes for isolated word speech recognition based on the combination of standard Hidden Markov Models (HMMs) and Complementary Gaussian Mixture Models (CGMMs) are proposed. Typically, in speech recognition systems, each word or phoneme in the vocabulary is represented by a model trained with samples of each particular class. The recognition is then performed ...

This paper describes the LIMSI Arabic Broadcast News system which produces a vowelized word transcription. The under 10x system, evaluated in the NIST RT-04F evaluation, uses a 3 pass decoding strategy with genderand bandwidth-specific acoustic models, a vowelized 65k word class pronunciation lexicon and a word-class 4-gram language model. In order to explicitly represent the vowelized word for...

The processing of single words that varied in their semantic (concrete/abstract word) and syntactic (content/function word) status was investigated under different task demands (semantic/ syntactic task) in an event-related functional magnetic resonance imaging experiment. Task demands to a large degree determined which subparts of the neuronal network supporting word processing were activated....

Purpose: The study aim was to determine whether the mean duration of silent pauses depends on the word class (open or closed) and to compare the duration for each type of word between children with typical language development and children with specific language impairment (SLI). Methods: The study included 40 children with typical language development and 20 children with SLI, aged between 7 a...

<div id="contentsContainer" class="style-scope qowt-page"><div id="contents" qowt-page"><p id="E129" class="x-scope qowt-word-para-4"><strong><span id="E130" class="qowt-font2-Arial">ABSTRAK </span></strong></p><p id="E131" qowt-word-para-5"><span id="E132" class="qowt-font2-Arial"><strong>Latar Belakang:</strong> </span...

Current speech recognition systems usually use word-based trigram language models. More elaborate models are applied to word lattices or N best lists in a rescoring pass following the acoustic decoding process. In this paper we consider techniques for dealing with class-based language models in the lattice rescoring framework of our JANUS large vocabulary speech recognizer. We demonstrate how t...

This paper proposes a new speech recognition scheme using three linguistic constraints. Multi-class composite bigram models [1] are used in the first and second passes to reflect word-neighboring characteristics as an extension of conventional word n-gram models. Trigram models with constituent boundary markers and word pattern models are both used in the third pass to utilize phrasal constrain...

In this paper, we describe an algorithm for aligning words with their translation in a bilingual corpus. Existing algorithms require enormous bilingual data to train statistical word-to-word translation models. Using word-based approach, frequent words with consistent translation can be aligned at a high precision rate. However, less frequent words or words with diverse translations usually do ...

Disambiguation Yorick Wilks and Mark Stevenson Department of Computer Science, University of She eld, Regent Court, 211 Portobello Street, She eld S1 4DP, UK fyorick, [email protected] Abstract Sense tagging, the automatic assignment of the appropriate sense from some lexicon to each of the words in a text, is a specialised instance of the general problem of word sense disambiguation. We di...