Abstract—Traditional-gram language models are widely used in state-of-the-art large vocabulary speech recognition systems. This simple model suffers from some limitations, such as overfitting of maximum-likelihood estimation and the lack of rich contextual knowledge sources. In this paper, we exploit a hierarchical Bayesian interpretation for language modeling, based on a nonparametric prior called Pitman–Yor process. This offers a principled approach to language model smoothing, embedding the power-law distribution for natural language. Experiments on the recognition of conversational speech in multiparty meetings demonstrate that by using hierarchical Bayesian language models, we are able to achieve significant reductions in perplexity and word error rate. Index Terms—AMI corpus, conversational speech recognition, hierarchical Bayesian model, language model (LM), meetings, smoothing. I.

The sequence memoizer is a model for sequence data with state-of-the-art performance on language modeling and compression. We propose a number of improvements to the model and inference algorithm, including an enlarged range of hyperparameters, a memory-efficient representation, and inference algorithms operating on the new representation. Our derivations are based on precise definitions of the various processes that will also allow us to provide an elementary proof of the “mysterious ” coagulation and fragmentation properties used in the original paper on the sequence memoizer by Wood et al. (2009). We present some experimental results supporting our improvements. 1

The Hierarchical Pitman Yor Process Language Model (HPYLM) is a Bayesian language model based on a nonparametric prior, the Pitman-Yor Process. It has been demonstrated, both theoretically and practically, that the HPYLM can provide better smoothing for language modeling, compared with state-of-the-art approaches such as interpolated Kneser-Ney and modified Kneser-Ney smoothing. However, estimation of Bayesian language models is expensive in terms of both computation time and memory; the inference is approximate and requires a number of iterations to converge. In this paper, we present a parallel training algorithm for the HPYLM, which enables the approach to be applied in the context of automatic speech recognition, using large training corpora with large vocabularies. We demonstrate the effectiveness of the proposed algorithm by estimating language models from corpora for meeting transcription containing over 200 million words, and observe significant reductions in perplexity and word error rate. Index Terms: language model, Pitman-Yor processes, hierarchical Bayesian models, parallel training, meetings

We present an unsupervised model for joint phrase alignment and extraction using nonparametric Bayesian methods and inversion transduction grammars (ITGs). The key contribution is that phrases of many granularities are included directly in the model through the use of a novel formulation that memorizes phrases generated not only by terminal, but also non-terminal symbols. This allows for a completely probabilistic model that is able to create a phrase table that achieves competitive accuracy on phrase-based machine translation tasks directly from unaligned sentence pairs. Experiments on several language pairs demonstrate that the proposed model matches the accuracy of traditional two-step word alignment/phrase extraction approach while reducing the phrase table to a fraction of the original size. 1

Frequency counts from very large corpora, such as the Web 1T dataset, have recently become available for language modeling. Omission of low frequency n-gram counts is a practical necessity for datasets of this size. Naive implementations of standard smoothing methods do not realize the full potential of such large datasets with missing counts. In this paper I present a new smoothing algorithm that combines the Dirichlet prior form of (Mackay and Peto, 1995) with the modified back-off estimates of (Kneser and Ney, 1995) that leads to a 31 % perplexity reduction on the Brown corpus compared to a baseline implementation of Kneser-Ney discounting. 1

We propose a novel dependent hierarchical Pitman-Yor process model for discrete data. An incremental Monte Carlo inference procedure for this model is developed. We show that inference in this model can be performed in constant space and linear time. The model is demonstrated in a discrete sequence prediction task where it is shown to achieve state of the art sequence prediction performance while using significantly less memory. 1.

Prior distributions play a crucial role in Bayesian approaches to clustering. Two commonly-used prior distributions are the Dirichlet and Pitman-Yor processes. In this paper, we investigate the predictive probabilities that underlie these processes, and the implicit “rich-get-richer ” characteristic of the resulting partitions. We explore an alternative prior for nonparametric Bayesian clustering—the uniform process—for applications where the “rich-get-richer ” property is undesirable. We also explore the cost of this process: partitions are no longer exchangeable with respect to the ordering of variables. We present new asymptotic and simulation-based results for the clustering characteristics of the uniform process and compare these with known results for the Dirichlet and Pitman-Yor processes. We compare performance on a real document clustering task, demonstrating the practical advantage of the uniform process despite its lack of exchangeability over orderings. 1

The hierarchical Pitman-Yor process-based smoothing method applied to language model was proposed by Goldwater and by Teh; the performance of this smoothing method is shown comparable with the modified Kneser-Ney method in terms of perplexity. Although this method was presented four years ago, there has been no paper which reports that this language model indeed improves translation quality in the context of Machine Translation (MT). This is important for the MT community since an improvement in perplexity does not always lead to an improvement in BLEU score; for example, the success of word alignment measured by Alignment Error Rate (AER) does not often lead to an improvement in BLEU. This paper reports in the context of MT that an improvement in perplexity really leads to an improvement in BLEU score. It turned out that an application of the Hierarchical Pitman-Yor Language Model (HPYLM) requires a minor change in the conventional decoding process. Additionally to this, we propose a new Pitman-Yor process-based statistical smoothing method similar to the Good-Turing method although the performance of this is inferior to HPYLM. We conducted experiments; HPYLM improved by 1.03 BLEU points absolute and 6 % relative for 50k EN-JP, which was statistically significant.

Using multi-layer neural networks to estimate the probabilities of word sequences is a promising research area in statistical language modeling, with applications in speech recognition and statistical machine translation. However, training such models for large vocabulary tasks is computationally challenging which does not scale easily to the huge corpora that are nowadays available. In this work, we study the performance and behavior of two neural statistical language models so as to highlight some important caveats of the classical training algorithms. The induced word embeddings for extreme cases are also analysed, thus providing insight into the convergence issues. A new initialization scheme and new training techniques are then introduced. These methods are shown to greatly reduce the training time and to significantly improve performance, both in terms of perplexity and on a large-scale translation task. 1

This paper provides an overview of nonparametric Bayesian models relevant to natural language processing (NLP) tasks. We first introduce Bayesian parametric methods, followed by nonparametric Bayesian modeling based on the most common nonparametric prior, the Dirichlet Process. We give characterizations of the Dirichlet Process via the Polya urn scheme, the related Chinese restaurant metaphor, and the stick-breaking construction. We will also introduce two generalizations