Statistical methods for speech recognition

By Frederick Jelinek. 1997

Subfields: Statistics and Probability, Speech & Audio

How researchers see this paper

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

In Statistics and Probability

In Speech & Audio

Citations in the corpus

A Theory of Multiple-Source Adaptation with Limited Target Labeled Data (2020, background): “…of the domain adaptation problem that arises in a variety of di↵erent applications, such as in natural language processing (Blitzer et al., 2007; Dredze et al., 2007; Jiang and Zhai, 2007), speech processing (Gauvain and Lee, 1994; Jelinek, 1997), and computer vision (Leggetter and Woodland, 1995).”
Agnostic Federated Learning (2019, background): “…Zhai, 2007; Raju et al., 2018), speech recognition (Legetter and Woodland, 1995; Gauvain and Chin-Hui, 1994; Pietra et al., 1992; Rosenfeld, 1996; Jelinek, 1998; Roark and Bacchiani, 2003), and computer vision (Mart´ınez, 2002) A problem more closely related to that of federated learning is that…”
Lightweight Probabilistic Deep Networks (2018, methodology): “The Dirichlet distribution only supports continuous vectors on the unit simplex; hence we render the discrete ground truth labels into continuous multinomial parameters by applying Laplace smoothing [26] with a small δ (1 e −3) to the labels.”
Domain Adaptation: Learning Bounds and Algorithms (2009, methodology): “ Jiang & Zhai, 2007; Chelba & Acero, 2006; Daume´ III & Marcu, 2006), speech processing (Legetter & Woodland, 1995; Gauvain & Chin-Hui, 1994; Pietra et al., 1992; Rosenfeld, 1996; Jelinek, 1998; Roark & Bacchiani, 2003), computer vision (Mart´ınez, 2002), and many other areas”
Stochastic Language Generation for Spoken Dialogue Systems (2000, methodology): “Using a smoothed n-gram would result in more randomness, but using the conventional back-off methods (Jelinek 1998), the probability mass assigned to unseen 5-grams would be very small, and those rare occurrences of unseen n-grams may not make sense anyway.”
Gmail Smart Compose: Real-Time Assisted Writing (2019, background): “• LogPerplexity is a typicalmeasure used for languagemodel evaluation [23].”
Multiple Source Adaptation and the Rényi Divergence (2009, background): “This problem arises in a variety of natu-ral language processing tasks such parsing, statistical language modeling, text classiﬁcation [15, 7, 16, 9, 10, 4, 6], or speech processing [11, 8, 14] and computer vision [13] tasks, as well as in many other applications.”
Kernel Methods for Relation Extraction (2002, methodology): “Again, we use the standard conﬁguration for both algorithms: for Naive Bayes we employed add-one smoothing (Jelinek, 1997); for Winnow, learning rate (promotion parameter) was set to 1.1 and the number of training set passes to 2.”
The Neural Noisy Channel (2016, background): “Noisy channel decompositions have been successfully used in a variety of problems, including speech recognition (Jelinek, 1998), machine translation (Brown et al.”
Cognitive Science in the era of Artificial Intelligence: A roadmap for reverse-engineering the infant language-learner (2016, positive): “…the next along with their probabilities or activation levels, enabling the errors and ambiguities to be resolved in a holistic and optimal fashion (for a statistical framework in speech processing, see Jelinek, 1997, for a review of natural language processing, see (J. H. Martin & Jurafsky, 2008)).”
Stochastic text generation (2000, methodology): “Standard techniques for nding paths through weighted lattices|as detailed, for example, in Jelinek’s (1997) textbook|are applicable to the reviewer’s path-decoding task.”
Estimation of statistical translation models based on mutual information for ad hoc information retrieval (2010, background): “In order to select words which are representative of a document, for each document d ∈ D , they compute the mutual information statistics [7] for each of its words according to: , where p ( w | d ) is the probability of word w in document d , and p ( w | D ) is the probability of word w in the…”
Dynamic Speech Models: Theory, Algorithms, and Applications (2006, methodology): “This new endeavor has been built upon the dramatic progress in speech technology achieved over the past two decades or so [10, 27–31].”
Learning with Mixtures of Trees (2000, methodology): “Chain graphs were introduced by Lauritzen (1996); they represent a superclass of both Bayesian networks and Markov random fields.”
Improving Offline Handwritten Text Recognition with Hybrid HMM/ANN Models (2011, unknown): “For offline handwritten text recognition, the image is converted into a sequence X ¼ ðx 1 . . . x m Þ of feature vectors and, under the statistical approach to pattern recognition [44], [45], the goal of general handwritten text recognition is to find the likeliest word sequence W ?”
Hierarchical Bayesian Language Models for Conversational Speech Recognition (2010, methodology): “Due to data sparsity, n-gram models based on maximum likelihood estimation (MLE) severely overfit the training data, resulting in unseen events being assigned zero probabilities [1].”
A new suffix tree similarity measure for document clustering (2007, background): “Bigrams, trigrams and much longer ngrams have been commonly used in statistical natural language processing [2, 9, 25].”