Connectionist Speech Recognition: A Hybrid Approach

By Hervé Bourlard, Nelson Morgan. 1993

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Natural Language Processing

Citations in the corpus

Speech recognition with deep recurrent neural networks (2013, background): “Neural networks have a long history in speech recognition, usually in combination with hidden Markov models [1, 2].”
HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Units (2021, methodology): “In the extreme case when α = 0, the loss is computed over the unmasked timesteps, which is similar to acoustic modeling in hybrid speech recognition systems [35]–[38].”
Unsupervised Speech Recognition (2021, methodology): “Such models are often classified into two categories: hybrid (Juang et al., 1986; Young, 1996; Povey, 2005; Hinton et al., 2012; Abdel-Hamid et al., 2012; Bourlard and Morgan, 2012) or end-to-end (Graves et al., 2006; Graves, 2012; Chorowski et al., 2015; Rao et al., 2017).”
Deep Speech 2 : End-to-End Speech Recognition in English and Mandarin (2015, methodology): “Feed-forward neural network acoustic models were explored more than 20 years ago [7, 50, 19].”
Recurrent Neural Network Regularization (2014, methodology): “Deep Neural Networks have been used for acoustic modeling for over half a century (see Bourlard & Morgan (1993) for a good review).”
Deep Speech: Scaling up end-to-end speech recognition (2014, background): “Neural network acoustic models and other connectionist approaches were ﬁrst introduced to speech pipelines in the early 1990s [1, 34, 11].”
Increasing the Interpretability of Recurrent Neural Networks Using Hidden Markov Models (2016, methodology): “M L ] 1 6 Ju n 20 16 ically train the RNN to predict HMM states (Bourlard & Morgan, 1994) or posteriors (Maas et al., 2012), referred to as hybrid or tandem methods respectively.”
Deep Learning: Methods and Applications (2014, methodology): “Note the use of “hybrid” in (3) above is different from that used sometimes in the literature, which refers to the hybrid systems for speech recognition feeding the output probabilities of a neural network into an HMM [17, 25, 42, 261].”
Syllable-Based Sequence-to-Sequence Speech Recognition with the Transformer in Mandarin Chinese (2018, background): “Sequence-to-sequence attention-based models integrate separate acoustic, pronunciation and language models of a conventional ASR system into a single neural network [5] and do not make the conditional independence assumptions as in standard hidden Markov based model [6].”
Improved training of end-to-end attention models for speech recognition (2018, methodology): “Conventional speech recognition systems [1] with neural network (NN) based acoustic models using the hybrid hidden Markov models (HMM) / NN approach [2, 3] usually operate on the phone level, given a phonetic pronunciation lexicon (from phones to words).”
Transformer-Based Acoustic Modeling for Hybrid Speech Recognition (2019, methodology): “% to 26% when the standard 4-gram language model (LM) is used; combined with neural LM rescoring, our system achieves state-of-the-art performance on this dataset. 2 Hybrid Architecture In hybrid ASR [22], an acoustic encoder is used to encode an input sequence x 1; ;x Tto a sequence of high level e”
Model Unit Exploration for Sequence-to-Sequence Speech Recognition (2019, background): “In fact, such models outperform conventional hybrid recognizers [13] when trained on sufficiently large amounts of data [9].”
Composition of Deep and Spiking Neural Networks for Very Low Bit Rate Speech Coding (2016, background): “The speech coding framework relies on a phonological (sub-phonetic) representation of speech.”
Large-Scale Deep Learning on the YFCC100M Dataset (2015, background): “Neural networks have been shown to perform well in a wide variety of tasks, including text analysis [4], speech recognition [5, 6, 7], various classiﬁcation tasks [8, 9], and most notably unsupervised and supervised feature learning on natural imagery [1, 2, 3].”
NeuralNetwork-Viterbi: A Framework for Weakly Supervised Video Learning (2018, methodology): “Since the outputs of the neural network are posterior probabilities p(c|xt), we follow the hybrid approach [3] and refactor”
A tutorial survey of architectures, algorithms, and applications for deep learning (2014, methodology): “Neural networks once were a popular approach but had not been competitive with the GMM–HMM [32, 33, 113, 129].”
Analysis of CNN-based speech recognition system using raw speech as input (2015, methodology): “In recent years, in the hybrid HMM/ANN framework [1], there has been growing interests in using “intermediate” representations instead of conventional features, such as cepstralbased features, as input for neural networks-based systems.”
Deep and Wide: Multiple Layers in Automatic Speech Recognition (2012, methodology): “REVIEW OF SOME RELEVANT EXPERIMENTS In a 1999 ICASSP paper [4], we empirically demonstrated the utility of using a sufficiently large hidden layer for three-layer networks used in the hybrid HMM/MLP system [13].”