Connectionist speech recognition of Broadcast News

Abstract

This paper describes connectionist techniques for recognition of Broadcast News. The fundamental difference between connectionist systems and more conventional mixture-of-Gaussian systems is that connectionist models directly estimate posterior probabilities as opposed to likelihoods. Access to posterior probabilities has enabled us to develop a number of novel approaches to confidence estimation, pronunciation modelling and search. In addition we have investigated a new feature extraction technique based on the modulationfiltered spectrogram, and methods for combining multiple information sources. We have incorporated all of these techniques into a system for the transcription of Broadcast News, and we present results on the 1998 DARPA Hub-4E Broadcast News evaluation data.

Context-Dependent Pre-Trained Deep Neural Networks for Large-Vocabulary Speech Recognition

“…A variety of different architectures and training algorithms have been proposed in the literature (see the comprehensive survey in [37]). Among these techniques, the ones most relevant to this work are those that use the ANNs to estimate the HMM stateposterior probabilities [38]- [45], which have been referred to as ANN-HMM hybrid models in the literature. In these ANN-HMM hybrid architectures, each output unit of the ANN is trained to estimate the posterior probability of a continuous density HMMs' state given the acoustic observations.…”

“…ANN-HMMs were later extended to model context-dependent phones and were applied to mid-vocabulary and some large vocabulary ASR tasks (e.g. in [45], which also employed recurrent neural architectures). However, in earlier work on context dependent ANN-HMM hybrid architectures [46], the posterior probability of the context-dependent phone was modeled as either…”

“…Much previous hybrid ANN-HMM work focused on context-independent or rudimentary context-dependent phone models and small to mid-vocabulary tasks (with notable exceptions such as [45]), possibly masking some of the potential advantages of the ANN-HMM hybrid approach. Additionally, GMM-HMM training is much easier to parallelize in a computer cluster setting, which historically gave such systems a significant advantage in scalability.…”

Automatic summarization of voicemail messages using lexical and prosodic features

“…This training set was augmented with those sentences from the Hub-4 Broadcast News and Switchboard language model training corpora which had a low perplexity with respect to the initial language model and the language model reestimated. We used a pronunciation dictionary containing around 10,000 words derived from the training data with pronunciations obtained from the SPRACH broadcast news system [Robinson et al 2002], plus 1,000 new words with pronunciations mainly constructed following the rules used to construct the broadcast news dictionary. The OOV rates were 1.6% on test42 and 2.0% on test50.…”

Multi-level adaptive networks in tandem and hybrid ASR systems

“…During the 1990s hybrid systems achieved good experimental results on some large vocabulary tasks [6,7]. However, HMM-GMM systems became increasingly more accurate owing to the use various techniques such as contextdependent phone modelling [8] speaker adaptation using MLLR [9], sequence-level discriminative training techniques such as MMI and MPE [10], and high-dimension feature space transforms such as fMPE [11], which either took advantage of the GMM structure or were much more computationally feasible for GMM-based systems compared with hybrid systems.…”