Continuous speech recognition by statistical methods

By Frederick Jelinek. 1976

How researchers see this paper

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

In Speech & Audio

Citations in the corpus

Natural Language Processing (Almost) from Scratch (2011, methodology): “For instance, modern speech recognition systems use Bayes rule to combine the outputs of an acoustic model trained on speech data and a language model trained on phonetic or textual corpora (Jelinek, 1976).”
Towards Unsupervised Speech Recognition and Synthesis with Quantized Speech Representation Learning (2019, background): “Hidden Markov Model performed on frame-level features were useful in early years, upon which various speech recognition [1] and synthesis [2] approaches were developed.”
HMM-Based Word Alignment in Statistical Translation (1996, methodology): “In our approach, we use a (cid:12)rst-order Hidden Markov model (HMM) (Jelinek, 1976), which is similar, but not identical to those used in speech recognition.”
Listening while speaking: Speech chain by deep learning (2017, methodology): “approaches have relied on acoustic-phonetics knowledge [2] in early works to template-based schemes with dynamic time warping (DTW) [3, 4] and data-driven approaches with rigorous statistical modeling of a hidden Markov model-Gaussian mixture model (HMM-GMM) [5, 6].”
A Nonparametric Bayesian Approach to Acoustic Model Discovery (2012, methodology): “HMM (θc) In our model, each HMM has three emission states, which correspond to the beginning, middle and end of a sub-word unit (Jelinek, 1976).”
The Application of Hidden Markov Models in Speech Recognition (2007, methodology): “…of modern HMM-based continuous speech recognition technology were laid down in the 1970's by groups at Carnegie- Mellon and IBM who introduced the use of discrete density HMMs 198 Introduction [11] [77] [108], and then later at Bell Labs [80] [81] [99] where continuous density HMMs were introduced.”
Connectionist sentence processing in perspective (1999, methodology): “We know from work on symbolic finite-state models such as Hidden Markov Models (HMM) and part-of-speech (POS) taggers (Jelinek, 1976; Church, 1988; Merialdo, 1994; Brill, 1992) that such approximations can achieve very high accuracy— better than 95% precision—without having any claim whatsoever to…”
Survey of the State of the Art in Human Language Technology (2010, background): “, Jelinek, F., Lafferty, J., Magerman, D. M., Mercer, D., and Roukos, S. (1993). Towards history-based grammars: Using richer models for probabilistic parsing.”
Tagging English Text with a Probabilistic Model (1994, methodology): “This maximization can be performed by the standard Forward-Backward (FB) or Baum-Welch algorithm (Baum and Eagon 1967; Jelinek 1976; Bahl, Jelinek, and Mercer 1983; Poritz 1988), by considering ~ and 1 - as the transition probabilities of a Markov model.”
Hybrid Autoregressive Transducer (HAT) (2020, methodology): “The automatic speech recognition (ASR) problem is probabilistically formulated as a maximum a posteriori decoding [1, 2]:”
Word Sense Disambiguation: The State of the Art (2005, methodology): “These methods were rediscovered first in speech processing (e.g. Jelinek, 1976; see the overview by Church and Mercer, 1993, and the collection of reprints by Waibel and Lee, 1990) and were immediately applied to written language analysis (e.g., in the work of Bahl and Mercer , 1976 ; Debili, 1977;…”
A stochastic model of human-machine interaction for learning dialog strategies (2000, background): “For instance, the tremendous progress made in speech recognition during the last decade is mainly due to the introduction and use of stochastic modeling of both acoustic [30] and linguistic phenomena [ 16 ].”
Unsupervised Lexicon Discovery from Acoustic Input (2015, background): “In particular, to capture the temporal dynamics of the features associated with a PLU, each HMM contains three emission states, which roughly correspond to the beginning, middle, and end of a phonetic unit (Jelinek, 1976).”
Discovering linguistic structures in speech: models and applications (2014, methodology): “states, which roughly correspond to the beginning, middle, and end of a phonetic unit [76].”
Automatic Speech Recognition - A Brief History of the Technology Development (2004, methodology): “While the difference in goals led to different realizations of the technology in various applications, the rapid development of statistical methods in the 1980’s, most notably the hidden Markov model (HMM) framework [34-35], caused a certain degree of convergence in the system design.”