N-gram posterior probability confidence measures for statistical machine translation: an empirical study

We report an empirical study of n-gram posterior probability confidence measures for statistical machine translation (SMT). We first describe an efficient and practical algorithm for rapidly computing n-gram posterior probabilities from large translation word lattices. These probabilities are shown to be a good predictor of whether or not the n-gram is found in human reference translations, motivating their use as a confidence measure for SMT. Comprehensive n-gram precision and word coverage measurements are presented for a variety of different language pairs, domains and conditions. We analyze the effect on reference precision of using single or multiple references, and compare the precision of posteriors computed from k-best lists to those computed over the full evidence space of the lattice. We also demonstrate improved confidence by combining multiple lattices in a multi-source translation framework.     

Backoff hierarchical class n-gram language models: effectiveness to model unseen events in speech recognition

In this paper, we introduce the backoff hierarchical class n-gram language models to better estimate the likelihood of unseen n-gram events. This multi-level class hierarchy language modeling approach generalizes the well-known backoff n-gram language modeling technique. It uses a class hierarchy to define word contexts. Each node in the hierarchy is a class that contains all the words of its descendant nodes. The closer a node to the root, the more general the class (and context) is. We investigate the effectiveness of the approach to model unseen events in speech recognition. Our results illustrate that the proposed technique outperforms backoff n-gram language models. We also study the effect of the vocabulary size and the depth of the class hierarchy on the performance of the approach. Results are presented on Wall Street Journal (WSJ) corpus using two vocabulary set: 5000 words and 20,000 words. Experiments with 5000 word vocabulary, which contain a small numbers of unseen events in the test set, show up to 10% improvement of the unseen event perplexity when using the hierarchical class n-gram language models. With a vocabulary of 20,000 words, characterized by a larger number of unseen events, the perplexity of unseen events decreases by 26%, while the word error rate (WER) decreases by 12% when using the hierarchical approach. Our results suggest that the largest gains in performance are obtained when the test set contains a large number of unseen events.     

利用语义词典Web挖掘语言模型的无指导译文消歧

为了解决困扰词义及译文消歧的数据稀疏及知识获取问题,提出一种基于Web利用n-gram统计语言模型进行消歧的方法.在提出词汇语义与其n-gram语言模型存在对应关系假设的基础上,首先利用Hownet建立中文歧义词的英文译文与知网DEF的对应关系并得到该DEF下的词汇集合,然后通过搜索引擎在Web上搜索,并以此计算不同DEF中词汇n-gram出现的概率,然后进行消歧决策.在国际语义评测SemEval-2007中的Multilingual Chinese English Lexical Sample Task测试集上的测试表明,该方法的P_mar值为55.9%,比其上该任务参评最好的无指导系统性能高出12.8%.     

Paraphrastic language models

Natural languages are known for their expressive richness. Many sentences can be used to represent the same underlying meaning. Only modelling the observed surface word sequence can result in poor context coverage and generalization, for example, when using n-gram language models (LMs). This paper proposes a novel form of language model, the paraphrastic LM, that addresses these issues. A phrase level paraphrase model statistically learned from standard text data with no semantic annotation is used to generate multiple paraphrase variants. LM probabilities are then estimated by maximizing their marginal probability. Multi-level language models estimated at both the word level and the phrase level are combined. An efficient weighted finite state transducer (WFST) based paraphrase generation approach is also presented. Significant error rate reductions of 0.5–0.6% absolute were obtained over the baseline n-gram LMs on two state-of-the-art recognition tasks for English conversational telephone speech and Mandarin Chinese broadcast speech using a paraphrastic multi-level LM modelling both word and phrase sequences. When it is further combined with word and phrase level feed-forward neural network LMs, a significant error rate reduction of 0.9% absolute (9% relative) and 0.5% absolute (5% relative) were obtained over the baseline n-gram and neural network LMs respectively.     

Language modelling for Russian and English using words and classes

This paper examines statistical language modelling of Russian and English in the context of automatic speech recognition. The characteristics of both a Russian and an English text corpus of similar composition are discussed with reference to the properties of both languages. In particular, it is shown that to achieve the same vocabulary coverage as a 65,000 word vocabulary for English, a 430,000 word vocabulary is required for Russian. The implications of this observation motivate the remainder of the paper. Perplexity experiments are reported for word-based N-gram modelling of the two languages and the differences are examined. It is found that, in contrast to English, there is little gain in using 4-grams over trigrams for modelling Russian. Class-based N-gram modelling is then considered and perplexity experiments are reported for two different types of class models, a two-sided model and a novel, one-sided model for which classes are generated automatically. Word and class model combinations show the two-sided model results in lower perplexities than combinations with the one-sided model. However, the very large Russian vocabulary favours the use of the one-sided model since the clustering algorithm, used to obtain word classes automatically, is significantly faster. Lattice rescoring experiments are then reported on an English-language broadcast news task which show that both combinations of the word model with either type of class model produce identical reductions in word error rate.     

Genericity and portability for task-independent speech recognition

As core speech recognition technology improves, opening up a wider range of applications, genericity and portability are becoming important issues. Most of todays recognition systems are still tuned to a particular task and porting the system to a new task (or language) requires a substantial investment of time and money, as well as human expertise.This paper addresses issues in speech recognizer portability and in the development of generic core speech recognition technology. First, the genericity of wide domain models is assessed by evaluating their performance on several tasks of varied complexity. Then, techniques aimed at enhancing the genericity of these wide domain models are investigated. Multi-source acoustic training is shown to reduce the performance gap between task-independent and task-dependent acoustic models, and for some tasks to out-perform task-dependent acoustic models.Transparent methods for porting generic models to a specific task are also explored. Transparent unsupervised acoustic model adaptation is contrasted with supervised adaptation, and incremental unsupervised adaptation of both the acoustic and linguistic models is investigated. Experimental results on a dialog task show that with the proposed scheme, a transparently adapted generic system can perform nearly as well (about a 1% absolute gap in word error rate) as a task-specific system trained on several tens of hours of manually transcribed data.     

Random forests and the data sparseness problem in language modeling

Language modeling is the problem of predicting words based on histories containing words already hypothesized. Two key aspects of language modeling are effective history equivalence classification and robust probability estimation. The solution of these aspects is hindered by the data sparseness problem.Application of random forests (RFs) to language modeling deals with the two aspects simultaneously. We develop a new smoothing technique based on randomly grown decision trees (DTs) and apply the resulting RF language models to automatic speech recognition. This new method is complementary to many existing ones dealing with the data sparseness problem. We study our RF approach in the context of n-gram type language modeling in which n − 1 words are present in a history. Unlike regular n-gram language models, RF language models have the potential to generalize well to unseen data, even when histories are longer than four words. We show that our RF language models are superior to the best known smoothing technique, the interpolated Kneser–Ney smoothing, in reducing both the perplexity (PPL) and word error rate (WER) in large vocabulary state-of-the-art speech recognition systems. In particular, we will show statistically significant improvements in a contemporary conversational telephony speech recognition system by applying the RF approach only to one of its many language models.     

Neural network language models for off-line handwriting recognition

Unconstrained off-line continuous handwritten text recognition is a very challenging task which has been recently addressed by different promising techniques. This work presents our latest contribution to this task, integrating neural network language models in the decoding process of three state-of-the-art systems: one based on bidirectional recurrent neural networks, another based on hybrid hidden Markov models and, finally, a combination of both. Experimental results obtained on the IAM off-line database demonstrate that consistent word error rate reductions can be achieved with neural network language models when compared with statistical N-gram language models on the three tested systems. The best word error rate, 16.1%, reported with ROVER combination of systems using neural network language models significantly outperforms current benchmark results for the IAM database.     

k-TSS language models in speech recognition systems

The aim of this work is to show the ability of stochastic regular grammars to generate accurate language models which can be well integrated, allocated and handled in a continuous speech recognition system. For this purpose, a syntactic version of the well-known n -gram model, called k -testable language in the strict sense (k -TSS), is used. The complete definition of a k -TSS stochastic finite state automaton is provided in the paper. One of the difficulties arising in representing a language model through a stochastic finite state network is that the recursive schema involved in the smoothing procedure must be adopted in the finite state formalism to achieve an efficient implementation of the backing-off mechanism. The use of the syntactic back-off smoothing technique applied to k -TSS language modelling allowed us to obtain a self-contained smoothed model integrating several k -TSS automata in a unique smoothed and integrated model, which is also fully defined in the paper. The proposed formulation leads to a very compact representation of the model parameters learned at training time: probability distribution and model structure. The dynamic expansion of the structure at decoding time allows an efficient integration in a continuous speech recognition system using a one-step decoding procedure. An experimental evaluation of the proposed formulation was carried out on two Spanish corpora. These experiments showed that regular grammars generate accurate language models (k -TSS) that can be efficiently represented and managed in real speech recognition systems, even for high values of k, leading to very good system performance.     

Graph vs. bag representation models for the topic classification of web documents

Text classification constitutes a popular task in Web research with various applications that range from spam filtering to sentiment analysis. In this paper, we argue that its performance depends on the quality of Web documents, which varies significantly. For example, the curated content of news articles involves different challenges than the user-generated content of blog posts and Social Media messages. We experimentally verify our claim, quantifying the main factors that affect the performance of text classification. We also argue that the established bag-of-words representation models are inadequate for handling all document types, as they merely extract frequent, yet distinguishing terms from the textual content of the training set. Thus, they suffer from low robustness in the context of noisy or unseen content, unless they are enriched with contextual, application-specific information. In their place, we propose the use of n-gram graphs, a model that goes beyond the bag-of-words representation, transforming every document into a graph: its nodes correspond to character or word n-grams and the co-occurring ones are connected by weighted edges. Individual document graphs can be combined into class graphs and graph similarities are employed to position and classify documents into the vector space. This approach offers two advantages with respect to bag models: first, classification accuracy increases due to the contextual information that is encapsulated in the edges of the n-gram graphs. Second, it reduces the search space to a limited set of robust, endogenous features that depend on the number of classes, rather than the size of the vocabulary. Our thorough experimental study over three large, real-world corpora confirms the superior performance of n-gram graphs across the main types of Web documents.     

Adaptation to non-native speech using evolutionary-based discriminative linear transforms

In this paper we are concerned with the problem of the adaptation of non-native speech in a large-vocabulary speech recognition system for Modern Standard Arabic (MSA). A technique to adapt Hidden Markov Models (HMMs) to foreign accents by using Genetic Algorithms (GAs) in unsupervised mode is presented. The implementation requirements of GAs, such as genetic operators and objective function, have been selected to give more reliability to a global linear transformation matrix. The Minimum Phone Error (MPE) criterion is used as an objective function. The West Point Language Data Consortium (LDC) modern standard Arabic database is used throughout our experiments. Results show that significant decrease of word error rate has been achieved by the evolutionary-based approach compared to conventional Maximum Likelihood Linear Regression (MLLR), Maximum a posteriori (MAP) techniques and to the adaptation combining MLLR and MPE-based training.     

HDP与互信息相结合的中文无指导分词

该文探讨了无指导条件下的中文分词,这对构建语言无关的健壮分词系统大有裨益。互信息与HDP(Hierarchical Dirichlet Process)是无指导情况下常用的分词模型,该文将两者结合,并改进了采样算法。不考虑标点符号,在两份大小不同的测试语料上获得的F值为0.693与0.741,相比baseline的HDP分别提升了5.8%和3.9%。该文还用该模型进行了半指导分词,实验结果比常用的CRF有指导分词提升了2.6%。     

EXPLOITING SYNTACTIC,SEMANTIC, AND LEXICAL REGULARITIES IN LANGUAGE MODELING VIA DIRECTED MARKOV RANDOM FIELDS

We present a directed Markov random field (MRF) model that combines n‐gram models, probabilistic context‐free grammars (PCFGs), and probabilistic latent semantic analysis (PLSA) for the purpose of statistical language modeling. Even though the composite directed MRF model potentially has an exponential number of loops and becomes a context‐sensitive grammar, we are nevertheless able to estimate its parameters in cubic time using an efficient modified Expectation‐Maximization (EM) method, the generalized inside–outside algorithm, which extends the inside–outside algorithm to incorporate the effects of the n‐gram and PLSA language models. We generalize various smoothing techniques to alleviate the sparseness of n‐gram counts in cases where there are hidden variables. We also derive an analogous algorithm to find the most likely parse of a sentence and to calculate the probability of initial subsequence of a sentence, all generated by the composite language model. Our experimental results on the Wall Street Journal corpus show that we obtain significant reductions in perplexity compared to the state‐of‐the‐art baseline trigram model with Good–Turing and Kneser–Ney smoothing techniques.     

Unsupervised Adaptation of Categorical Prosody Models for Prosody Labeling and Speech Recognition

Automatic speech recognition (ASR) systems rely almost exclusively on short-term segment-level features (MFCCs), while ignoring higher level suprasegmental cues that are characteristic of human speech. However, recent experiments have shown that categorical representations of prosody, such as those based on the Tones and Break Indices (ToBI) annotation standard, can be used to enhance speech recognizers. However, categorical prosody models are severely limited in scope and coverage due to the lack of large corpora annotated with the relevant prosodic symbols (such as pitch accent, word prominence, and boundary tone labels). In this paper, we first present an architecture for augmenting a standard ASR with symbolic prosody. We then discuss two novel, unsupervised adaptation techniques for improving, respectively, the quality of the linguistic and acoustic components of our categorical prosody models. Finally, we implement the augmented ASR by enriching ASR lattices with the adapted categorical prosody models. Our experiments show that the proposed unsupervised adaptation techniques significantly improve the quality of the prosody models; the adapted prosodic language and acoustic models reduce binary pitch accent (presence versus absence) classification error rate by 13.8% and 4.3%, respectively (relative to the seed models) on the Boston University Radio News Corpus, while the prosody-enriched ASR exhibits a 3.1% relative reduction in word error rate (WER) over the baseline system.     

A New Multiword Expression Metric and Its Applications

Multiword Expressions(MWEs) appear frequently and ungrammatically in natural languages.Identifying MWEs in free texts is a very challenging problem.This paper proposes a knowledge-free,unsupervised,and language-independent Multiword Expression Distance(MED).The new metric is derived from an accepted physical principle,measures the distance from an n-gram to its semantics,and outperforms other state-of-the-art methods on MWEs in two applications: question answering and named entity extraction.     

Finding words in alphabet soup: Inference on freeform character recognition for historical scripts

This paper develops word recognition methods for historical handwritten cursive and printed documents. It employs a powerful segmentation-free letter detection method based upon joint boosting with histograms of gradients as features. Efficient inference on an ensemble of hidden Markov models can select the most probable sequence of candidate character detections to recognize complete words in ambiguous handwritten text, drawing on character n-gram and physical separation models. Experiments with two corpora of handwritten historic documents show that this approach recognizes known words more accurately than previous efforts, and can also recognize out-of-vocabulary words.