The effect of language models on phonetic decoding for spoken term detection

Spoken term detection (STD) popularly involves performing word or sub-word level speech recognition and indexing the result. This work challenges the assumption that improved speech recognition accuracy implies better indexing for STD. Using an index derived from phone lattices, this paper examines the effect of language model selection on the relationship between phone recognition accuracy and STD accuracy. Results suggest that language models usually improve phone recognition accuracy but their inclusion does not always translate to improved STD accuracy. The findings suggest that using phone recognition accuracy to measure the quality of an STD index can be problematic, and highlight the need for an alternative that is more closely aligned with the goals of the specific detection task.

High-Order Concept Associations Mining and Inferential Language Modeling for Online Review Spam Detection

Despite many incidents about fake online consumer reviews have been reported, very few studies have been conducted to date to examine the trustworthiness of online consumer reviews. One of the reasons is the lack of an effective computational method to separate the untruthful reviews (i.e., spam) from the legitimate ones (i.e., ham) given the fact that prominent spam features are often missing in online reviews. The main contribution of our research work is the development of a novel review spam detection method which is underpinned by an unsupervised inferential language modeling framework. Another contribution of this work is the development of a high-order concept association mining method which provides the essential term association knowledge to bootstrap the performance for untruthful review detection. Our experimental results confirm that the proposed inferential language model equipped with high-order concept association knowledge is effective in untruthful review detection when compared with other baseline methods.

Text mining and probabilistic language modeling for online review spam detecting

In the era of Web 2.0, huge volumes of consumer reviews are posted to the Internet every day. Manual approaches to detecting and analyzing fake reviews (i.e., spam) are not practical due to the problem of information overload. However, the design and development of automated methods of detecting fake reviews is a challenging research problem. The main reason is that fake reviews are specifically composed to mislead readers, so they may appear the same as legitimate reviews (i.e., ham). As a result, discriminatory features that would enable individual reviews to be classified as spam or ham may not be available. Guided by the design science research methodology, the main contribution of this study is the design and instantiation of novel computational models for detecting fake reviews. In particular, a novel text mining model is developed and integrated into a semantic language model for the detection of untruthful reviews. The models are then evaluated based on a real-world dataset collected from amazon.com. The results of our experiments confirm that the proposed models outperform other well-known baseline models in detecting fake reviews. To the best of our knowledge, the work discussed in this article represents the first successful attempt to apply text mining methods and semantic language models to the detection of fake consumer reviews. A managerial implication of our research is that firms can apply our design artifacts to monitor online consumer reviews to develop effective marketing or product design strategies based on genuine consumer feedback posted to the Internet.

Facilitating query decomposition in query language modeling by association rule mining using multiple sliding windows

This paper presents a novel framework to further advance the recent trend of using query decomposition and high-order term relationships in query language modeling, which takes into account terms implicitly associated with different subsets of query terms. Existing approaches, most remarkably the language model based on the Information Flow method are however unable to capture multiple levels of associations and also suffer from a high computational overhead. In this paper, we propose to compute association rules from pseudo feedback documents that are segmented into variable length chunks via multiple sliding windows of different sizes. Extensive experiments have been conducted on various TREC collections and our approach significantly outperforms a baseline Query Likelihood language model, the Relevance Model and the Information Flow model.

Topic dependent language modelling for spoken term detection

This paper investigates the effect of topic dependent language models (TDLM) on phonetic spoken term detection (STD) using dynamic match lattice spotting (DMLS). Phonetic STD consists of two steps: indexing and search. The accuracy of indexing audio segments into phone sequences using phone recognition methods directly affects the accuracy of the final STD system. If the topic of a document in known, recognizing the spoken words and indexing them to an intermediate representation is an easier task and consequently, detecting a search word in it will be more accurate and robust. In this paper, we propose the use of TDLMs in the indexing stage to improve the accuracy of STD in situations where the topic of the audio document is known in advance. It is shown that using TDLMs instead of the traditional general language model (GLM) improves STD performance according to figure of merit (FOM) criteria.

Developing and testing readability measurements for second language learners

This research constructed a readability measurement for French speakers who view English as a second language. It identified the true cognates, which are the similar words from these two languages, as an indicator of the difficulty of an English text for French people. A multilingual lexical resource is used to detect true cognates in text, and Statistical Language Modelling to predict the predict the readability level. The proposed enhanced statistical language model is making a step in the right direction by improving the accuracy of readability predictions for French speakers by up to 10% compared to state of the art approaches. The outcome of this study could accelerate the learning process for French speakers who are studying English. More importantly, this study also benefits the readability estimation research community, presenting an approach and evaluation at sentence level as well as innovating with the use of cognates as a new text feature.

Language identification using parallel sub-word recognition

Parallel sub-word recognition (PSWR) is a new model that has been proposed for language identification (LID) which does not need elaborate phonetic labeling of the speech data in a foreign language. The new approach performs a front-end tokenization in terms of sub-word units which are designed by automatic segmentation, segment clustering and segment HMM modeling. We develop PSWR based LID in a framework similar to the parallel phone recognition (PPR) approach in the literature. This includes a front-end tokenizer and a back-end language model, for each language to be identified. Considering various combinations of the statistical evaluation scores, it is found that PSWR can perform as well as PPR, even with broad acoustic sub-word tokenization, thus making it an efficient alternative to the PPR system.

Suite of tools for statistical N-gram language modeling for pattern mining in whole genome sequences

Genome sequences contain a number of patterns that have biomedical significance. Repetitive sequences of various kinds are a primary component of most of the genomic sequence patterns. We extended the suffix-array based Biological Language Modeling Toolkit to compute n-gram frequencies as well as n-gram language-model based perplexity in windows over the whole genome sequence to find biologically relevant patterns. We present the suite of tools and their application for analysis on whole human genome sequence.

Language models for online handwritten Tamil word recognition

N-gram language models and lexicon-based word-recognition are popular methods in the literature to improve recognition accuracies of online and offline handwritten data. However, there are very few works that deal with application of these techniques on online Tamil handwritten data. In this paper, we explore methods of developing symbol-level language models and a lexicon from a large Tamil text corpus and their application to improving symbol and word recognition accuracies. On a test database of around 2000 words, we find that bigram language models improve symbol (3%) and word recognition (8%) accuracies and while lexicon methods offer much greater improvements (30%) in terms of word recognition, there is a large dependency on choosing the right lexicon. For comparison to lexicon and language model based methods, we have also explored re-evaluation techniques which involve the use of expert classifiers to improve symbol and word recognition accuracies.

Paraphrastic language models

In natural languages multiple word sequences can represent the same underlying meaning. Only modelling the observed surface word sequence can result in poor context coverage, for example, when using n-gram language models (LM). To handle this issue, this paper presents a novel form of language model, the paraphrastic LM. A phrase level transduction model that is statistically learned from standard text data is used to generate paraphrase variants. LM probabilities are then estimated by maximizing their marginal probability. Significant error rate reductions of 0.5%-0.6% absolute were obtained on a state-ofthe-art conversational telephone speech recognition task using a paraphrastic multi-level LM modelling both word and phrase sequences.

Smoothing LDA Model for Text Categorization

Abstract. Latent Dirichlet Allocation (LDA) is a document level language model. In general, LDA employ the symmetry Dirichlet distribution as prior of the topic-words’ distributions to implement model smoothing. In this paper, we propose a data-driven smoothing strategy in which probability mass is allocated from smoothing-data to latent variables by the intrinsic inference procedure of LDA. In such a way, the arbitrariness of choosing latent variables'priors for the multi-level graphical model is overcome. Following this data-driven strategy,two concrete methods, Laplacian smoothing and Jelinek-Mercer smoothing, are employed to LDA model. Evaluations on different text categorization collections show data-driven smoothing can significantly improve the performance in balanced and unbalanced corpora.

Mutual Information and Perplexity based Clustering of Dialogue Information for Dynamic Adaptation of Language Models

We present two approaches to cluster dialogue-based information obtained by the speech understanding module and the dialogue manager of a spoken dialogue system. The purpose is to estimate a language model related to each cluster, and use them to dynamically modify the model of the speech recognizer at each dialogue turn. In the first approach we build the cluster tree using local decisions based on a Maximum Normalized Mutual Information criterion. In the second one we take global decisions, based on the optimization of the global perplexity of the combination of the cluster-related LMs. Our experiments show a relative reduction of the word error rate of 15.17%, which helps to improve the performance of the understanding and the dialogue manager modules.

Mutual Information and Perplexity based Clustering of Dialogue Information for Dynamic Adaptation of Language Models

We present two approaches to cluster dialogue-based information obtained by the speech understanding module and the dialogue manager of a spoken dialogue system. The purpose is to estimate a language model related to each cluster, and use them to dynamically modify the model of the speech recognizer at each dialogue turn. In the first approach we build the cluster tree using local decisions based on a Maximum Normalized Mutual Information criterion. In the second one we take global decisions, based on the optimization of the global perplexity of the combination of the cluster-related LMs. Our experiments show a relative reduction of the word error rate of 15.17%, which helps to improve the performance of the understanding and the dialogue manager modules.

Language identification based on a discriminative text categorization technique

In this paper, we describe new results and improvements to a lan-guage identification (LID) system based on PPRLM previously introduced in [1] and [2]. In this case, we use as parallel phone recognizers the ones provided by the Brno University of Technology for Czech, Hungarian, and Russian lan-guages, and instead of using traditional n-gram language models we use a lan-guage model that is created using a ranking with the most frequent and discrim-inative n-grams. In this language model approach, the distance between the ranking for the input sentence and the ranking for each language is computed, based on the difference in relative positions for each n-gram. This approach is able to model reliably longer span information than in traditional language models obtaining more reliable estimations. We also describe the modifications that we have being introducing along the time to the original ranking technique, e.g., different discriminative formulas to establish the ranking, variations of the template size, the suppression of repeated consecutive phones, and a new clus-tering technique for the ranking scores. Results show that this technique pro-vides a 12.9% relative improvement over PPRLM. Finally, we also describe re-sults where the traditional PPRLM and our ranking technique are combined.

Dynamic topic-based adaptation of language models: a comparison between different approaches

This paper presents a dynamic LM adaptation based on the topic that has been identified on a speech segment. We use LSA and the given topic labels in the training dataset to obtain and use the topic models. We propose a dynamic language model adaptation to improve the recognition performance in "a two stages" AST system. The final stage makes use of the topic identification with two variants: the first on uses just the most probable topic and the other one depends on the relative distances of the topics that have been identified. We perform the adaptation of the LM as a linear interpolation between a background model and topic-based LM. The interpolation weight id dynamically adapted according to different parameters. The proposed method is evaluated on the Spanish partition of the EPPS speech database. We achieved a relative reduction in WER of 11.13% over the baseline system which uses a single blackground LM.