A goal of statistical language modeling is to learn the joint probability function of sequences of words in a language. This is intrinsically difficult because of the curse of dimensionality: a word sequence on which the model will be tested is likely to be different from all the word sequences seen during training. Traditional but very successful approaches based on n-grams obtain generalization by concatenating very short overlapping sequences seen in the training set. We propose to fight the curse of dimensionality by learning a distributed representation for words which allows each training sentence to inform the model about an exponential number of semantically neighboring sentences. The model learns simultaneously (1) a distributed representation for each word along with (2) the probability function for word sequences, expressed in terms of these representations. Generalization is obtained because a sequence of words that has never been seen before gets high probability if it is made of words that are similar (in the sense of having a nearby representation) to words forming an already seen sentence. Training such large models (with millions of parameters) within a reasonable time is itself a significant challenge. We report on experiments using neural networks for the probability function, showing on two text corpora that the proposed approach significantly improves on state-of-the-art n-gram models, and that the proposed approach allows to take advantage of longer contexts.

/pdf/a-neural-probabilistic-language-model-v51mwoff5f.pdf

A neural probabilistic language model

In recent years, variants of a neural network architecture for statistical language modeling have been proposed and successfully applied, e.g. in the language modeling component of speech recognizers. The main advantage of these architectures is that they learn an embedding for words (or other symbols) in a continuous space that helps to smooth the language model and provide good generalization even when the number of training examples is insufficient. However, these models are extremely slow in comparison to the more commonly used n-gram models, both for training and recognition. As an alternative to an importance sampling method proposed to speed-up training, we introduce a hierarchical decomposition of the conditional probabilities that yields a speed-up of about 200 both during training and recognition. The hierarchical decomposition is a binary hierarchical clustering constrained by the prior knowledge extracted from the WordNet semantic hierarchy.

/pdf/hierarchical-probabilistic-neural-network-language-model-2469j48wr2.pdf

Hierarchical Probabilistic Neural Network Language Model.

Hidden Markov Models (HMMs) provide a simple and effective framework for modelling time-varying spectral vector sequences. As a consequence, almost all present day large vocabulary continuous speech recognition (LVCSR) systems are based on HMMs.

Whereas the basic principles underlying HMM-based LVCSR are rather straightforward, the approximations and simplifying assumptions involved in a direct implementation of these principles would result in a system which has poor accuracy and unacceptable sensitivity to changes in operating environment. Thus, the practical application of HMMs in modern systems involves considerable sophistication.

The aim of this review is first to present the core architecture of a HMM-based LVCSR system and then describe the various refinements which are needed to achieve state-of-the-art performance. These refinements include feature projection, improved covariance modelling, discriminative parameter estimation, adaptation and normalisation, noise compensation and multi-pass system combination. The review concludes with a case study of LVCSR for Broadcast News and Conversation transcription in order to illustrate the techniques described.

/pdf/application-of-hidden-markov-models-in-speech-recognition-4ic3ad3d5g.pdf

Application of Hidden Markov Models in Speech Recognition

A central goal of statistical language modeling is to learn the joint probability function of sequences of words in a language. This is intrinsically difficult because of the curse of dimensionality: a word sequence on which the model will be tested is likely to be different from all the word sequences seen during training. Traditional but very successful approaches based on n-grams obtain generalization by concatenating very short overlapping sequences seen in the training set. We propose to fight the curse of dimensionality by learning a distributed representation for words which allows each training sentence to inform the model about an exponential number of semantically neighboring sentences. Generalization is obtained because a sequence of words that has never been seen before gets high probability if it is made of words that are similar (in the sense of having a nearby representation) to words forming an already seen sentence. Training such large models (with millions of parameters) within a reasonable time is itself a significant challenge. We report on several methods to speed-up both training and probability computation, as well as comparative experiments to evaluate the improvements brought by these techniques. We finally describe the incorporation of this new language model into a state-of-the-art speech recognizer of conversational speech.

Neural Probabilistic Language Models

In the past several years, a number of different language modeling improvements over simple trigram models have been found, including caching, higher-order n-grams, skipping, interpolated Kneser-Ney smoothing, and clustering. We present explorations of variations on, or of the limits of, each of these techniques, including showing that sentence mixture models may have more potential. While all of these techniques have been studied separately, they have rarely been studied in combination. We find some significant interactions, especially with smoothing and clustering techniques. We compare a combination of all techniques together to a Katz smoothed trigram model with no count cutoffs. We achieve perplexity reductions between 38% and 50% (1 bit of entropy), depending on training data size, as well as a word error rate reduction of 8.9%. Our perplexity reductions are perhaps the highest reported compared to a fair baseline. This is the extended version of the paper; it contains additional details and proofs, and is designed to be a good introduction to the state of the art in language modeling.

A Bit of Progress in Language Modeling

This paper describes the 1998 HTK large vocabulary speech recognition system for conversational telephone speech as used in the NIST 1998 Hub5E evaluation. Front-end and language modelling experiments conducted using various training and test sets from both the Switchboard and Callhome English corpora are presented. Our complete system includes reduced bandwidth analysis, side-based cepstral feature normalisation, vocal tract length normalisation (VTLN), triphone and quinphone hidden Markov models (HMMs) built using speaker adaptive training (SAT), maximum likelihood linear regression (MLLR) speaker adaptation and a confidence score based system combination. A detailed description of the complete system together with experimental results for each stage of our multi-pass decoding scheme is presented. The word error rate obtained is almost 20% better than our 1997 system on the development set.

/pdf/the-1998-htk-system-for-transcription-of-conversational-13mdka5rj0.pdf

The 1998 HTK system for transcription of conversational telephone speech

This paper compares various category-based language models when used in conjunction with a word-based trigram by means of linear interpolation. Categories corresponding to parts-of-speech as well as automatically clustered groupings are considered. The category-based model employs variable-length n-grams and permits each word to belong to multiple categories. Relative word error rate reductions of between 2 and 7% over the baseline are achieved in N-best rescoring experiments on the Wall Street Journal corpus. The largest improvement is obtained with a model using automatically determined categories. Perplexities continue to decrease as the number of different categories is increased, but improvements in the word error rate reach an optimum.

Comparison of part-of-speech and automatically derived category-based language models for speech recognition

This paper presents the recent development of the HTK broadcast news transcription system. Previously we have used data type specific modelling based on adapted Wall Street Journal trained HMMs. However, we are now using data for which no manual preclassification or segmentation is available and therefore automatic techniques are required and compatible acoustic modelling strategies must be adopted. A number of recognition experiments are presented that compare data-type specific and non-specific models; differing amounts of training data; the use of gender-dependent modelling and the effects of automatic data-type classification. Based on these experiments, the HTK system for the 1997 broadcast news evaluation was designed. A detailed description of this system is given which includes a class-based language modelling component. The complete system yields an overall word error rate of 22.0% on the 1996 unpartitioned broadcast news development test data and just 15.8% on the 1997 evaluation test set.

/pdf/the-1997-htk-broadcast-news-transcription-system-yq3kr52ctb.pdf

The 1997 HTK broadcast news transcription system

This paper presents the development of the HTK broadcast news transcription system for the November 1998 Hub4 evaluation. Relative to the previous year’s system The system a number of features were added including vocal tract length normalisation; cluster-based variance normalisation; double the quantity of acoustic training data; interpolated word level language models to combine text sources; increased broadcast news language model training data; and an extra adaptation stage using a full-variance transform. Overall these changes to the system reduced the error rate by 13% on the 1997 evaluation data and the final system had an overall word error rate of 13.8% for the 1998 evaluation data sets.

/pdf/the-1998-htk-broadcast-news-transcription-system-development-59snj8xo12.pdf

The 1998 HTK broadcast news transcription system: development and results

Investigates the perplexity and word error rate performance of two different forms of class model and the respective data-driven algorithms for obtaining automatic word classifications. The computational complexity of the algorithm for the 'conventional' two-sided class model is found to be unsuitable for very large vocabularies (>100k) or large numbers of classes (>2000). A one-sided class model is therefore investigated and the complexity of its algorithm is found to be substantially less in such situations. Perplexity results are reported on both English and Russian data. For the latter both 65k and 430k vocabularies are used. Lattice rescoring experiments are also performed on an English language broadcast news task. These experimental results show that both models, when interpolated with a word model, perform similarly well. Moreover, classifications are obtained for the one-sided model in a fraction of the time required by the two-sided model, especially for very large vocabularies.

/pdf/efficient-class-based-language-modelling-for-very-large-3y0at92o8w.pdf

E.W.D. Whittaker

Papers

The 1998 HTK system for transcription of conversational telephone speech

Comparison of part-of-speech and automatically derived category-based language models for speech recognition

The 1997 HTK broadcast news transcription system

The 1998 HTK broadcast news transcription system: development and results

Efficient class-based language modelling for very large vocabularies