Search

Identifying the language of a text is an important step for several natural language processing applications. State-of-the-art language identification (LID) systems perform very well when discriminating between unrelated languages on standard datasets. However, the LID task has a bottleneck when discriminating between similar languages or language varieties. Furthermore, LID has also proven to be very challenging when dealing with short texts such as the ones from Twitter. In this paper, we propose the use of smoothed n-gram language models to classify tweets in both Brazilian and European Portuguese variants. Word and character n-gram language models were combined and evaluated through five different classifiers. We have compared the smoothed n-gram language models together with the Term Frequency and Inverse Document Frequency weighting scheme. This paper also proposes an ensemble model, in which the class labels output were combined using majority voting and algebraic combiners. The best configuration reached accuracy of 92.71% using an ensemble model, which combines Lidstone (0.1) character 6-gram, Good–Turing word unigram, and Witten–Bell word bigram models, together with the Log-Likelihood Ratio estimation method.

Humans form concept of objects by classifying them into categories, and acquire language by simultaneously interacting with others. Thus, the meaning of a word can be learned by connecting a recognized word to its corresponding concept. We consider this ability important for robots to flexibly develop knowledge of language and concepts. In this paper, we propose an online algorithm for robots to acquire knowledge of natural language and learn object concepts. A robot learns the language model from word sequences, which are obtained by the segmentation of phoneme sequences provided by a user, by using unsupervised word segmentation each time it is provided with a new object. Moreover, the robot acquires object concepts using these word sequences as well as multimodal information obtained by observing objects. The crucial aspect of our model is the interdependence of words and concepts: there is a high probability that the same words will be uttered to describe objects in the same category. By taking this relationship into account, our proposed method enables robots to acquire a more accurate language model and object concepts online. Experimental results verify this.

Speech emotion recognition continues to attract a lot of research especially under mixed-language scenarios. Here, we show that emotion is language dependent and that enhanced emotion recognition systems can be built when the language is known. We propose a two-stage emotion recognition system that starts by identifying the language, followed by a dedicated language-dependent recognition system for identifying the type of emotion. The system is able to recognize accurately the four main types of emotion, namely neutral, happy, angry, and sad. These types of emotion states are widely used in practical setups. To keep the computation complexity low, we identify the language using a feature vector consisting of energies from a basic wavelet decomposition. A hidden Markov model (HMM) is then used to track the changes of this vector to identify the language, achieving recognition accuracy close to 100%. Once the language is identified, a set of speech processing features including pitch and MFCCs are used with a neural network (NN) architecture to identify the emotion type. The results show that that identifying the language first can substantially improve the overall accuracy in identifying emotions. The overall accuracy achieved with the proposed system reached more than 93%. To test the robustness of the proposed methodology, we also used a Gaussian mixture model (GMM) for both language identification and emotion recognition. Our proposed HMM-NN approach showed a better performance than the GMM-based approach. More importantly, we tested the proposed algorithm with 6 emotions which are showed that the overall accuracy continues to be excellent, while the performance of the GMM-based approach deteriorates substantially. It is worth noting that the performance we achieved is close to the one attained for single language emotion recognition systems and outperforms by far recognition systems without language identification (around 60%). The work shows the strong correlation between language and type of emotion, and can further be extended to other scenarios including gender-based, facial expression-based, and age-based emotion recognition.

Many of the world's around 6,000 languages are in danger of disappearing as people give up use of a minority language in favor of the majority language in a process called language shift. Language shift can be monitored on a large scale through the use of mathematical models by way of differential equations, for example, reaction-diffusion equations. Here, we use a different approach: we propose a model for language dynamics based on the principles of cellular automata/agent-based modeling and combine it with very detailed empirical data. Our model makes it possible to follow language dynamics over space and time, whereas existing models based on differential equations average over space and consequently provide no information on local changes in language use. Additionally, cellular automata models can be used even in cases where models based on differential equations are not applicable, for example, in situations where one language has become dispersed and retreated to language islands. Using data from a bilingual region in Austria, we show that the most important factor in determining the spread and retreat of a language is the interaction with speakers of the same language. External factors like bilingual schools or parish language have only a minor influence.

Hidden Markov Model is an empirical tool that can be used in many applications related to natural language processing. In this paper a comparative study was conducted between different applications in natural Arabic language processing that uses Hidden Markov Model such as morphological analysis, part of speech tagging, text classification, and name entity recognition. Comparative results showed that HMM can be used in different layers of natural language processing, but mainly in pre-processing phase such as: part of speech tagging, morphological analysis and syntactic structure; however in high level applications text classification their use is limited to certain number of researches.

The objective of Native Language Identification is to determine the native language of the author of a text that he or she wrote in another language. By contrast, Language Variety Identification aims at classifying texts representing different varieties of a single language. We postulate that both tasks may be reduced to a single objective, which is to identify the language variety of the text. We design a general approach that combines string kernels and word embeddings, which capture different characteristics of texts. The results of our experiments show that the approach achieves excellent results on both tasks, without any task-specific adaptations.

While Application Programming Interface (API) enables an easy and flexible software development process, selecting a best-fit API is often non-straightforward in practice due to misunderstanding on the API specification or a complex programming context etc.. Consequently, the API selection has always been time-consuming and error-prone. In recent years, API recommendation systems have been introduced to help developers choose an API automatically, e.g., Eclipse and IntelliJ can generate an internal or user-defined API on the fly. Other research leveraged language models to capture the regularity in API usage and further guide the completion of APIs. While existing approaches provided a general support for API usage, they suffer from the lack of semantic awareness (e.g., Eclipse) and code relevance (e.g., language model based methods). To overcome these limitations, we proposed Crmac in this paper. The key insight of Crmac is a combination of a cache language model which learns code regularity from both open-source projects and local projects, as well as a relevance mining engine that identifies similar code to enable a weighted language model training. In our empirical evaluation, Crmac overwhelmed n-gram approaches, with an improvement of 5.28% in terms of top 10 accuracy. Moreover, over 79% APIs were correctly predicted in the top 10 guesses of Crmac.

International audience; Several models have proposed that language control occurs between language representations, such as language tags, and between lemmas. Yet, most research has solely focused on language-control processes between language representations. In the present study, we investigated whether language control can also occur between lemmas by allowing bilinguals to practice a language or language-specific items prior to a language-switching task, and thus change the relative activation of the language representations and/or lemmas. By changing the activation levels, relatively more language control should occur for this language representation and/or lemma relative to their translation-equivalent due to the reactive nature of language control. The results showed that this was all the more so when language-specific items were practiced than when merely a language was practiced. Hence, the current study provides evidence that language control is not restricted to language representations, but could also occur between lemmas.

Language identification is widely used in machine learning, text mining, information retrieval, and speech processing. Available techniques for solving the problem of language identification do require large amount of training text that are not available for under-resourced languages which form the bulk of the World’s languages. The primary objective of this study is to propose a lexicon based algorithm which is able to perform language identification using minimal training data. Because language identification is often the first step in many natural language processing tasks, it is necessary to explore techniques that will perform language identification in the shortest possible time. Hence, the second objective of this research is to study the effect of the proposed algorithm on the run-time performance of language identification. Precision, recall, and F1 measures were used to determine the effectiveness of the proposed word length algorithm using datasets drawn from the Universal Declaration of Human Rights Act in 15 languages. The experimental results show good accuracy on language identification at the document level and at the sentence level based on the available dataset. The improved algorithm also showed significant improvement in run time performance compared with the spelling checker approach.

Language is usually modelled through a predominantly synoptic perspective; even if the object of analysis is spoken language, we tend to look at extracts where the analysis of parts makes use of the whole. Holistic analyses can be very good for capturing realities of language in many respects, but in the case of modelling temporal aspects of processing they fall into the trap of unrealistic hindsight. The experience of speech is time-bound: a hearer will go on what he or she has heard at any given point, and will anticipate what may follow. The predictions will be either confirmed or rejected in rapid succession, as speech moves on. The time window for working memory is very brief, and processing focuses on continuously changing input. Models of this process must take into account this dynamism, and they need to take on board the fact that language must be continually processed even while utterances are still incomplete. Most models of language structure are based on completed units; this tends to lead to a hierarchical view of language, embodied in most grammars. The reality of temporally progressing speech is nevertheless fundamentally linear along the dimension of time.

The language model is a widely used component in fields such as natural language processing, automatic speech recognition, and optical character recognition. In particular, statistical machine translation uses language models, and the translation speed and the amount of memory required are greatly affected by the performance of the language model implementation. We propose a fast and compact implementation of n -gram language models that increases query speed and reduces memory usage by using a double-array structure, which is known to be a fast and compact trie data structure. We propose two types of implementation: one for backward suffix trees and the other for reverse tries. The data structure is optimized for space efficiency by embedding model parameters into otherwise unused spaces in the double-array structure. We show that the reverse trie version of our method is among the smallest state-of-the-art implementations in terms of model size with almost the same speed as the implementation that performs fastest on perplexity calculation tasks. Similarly, we achieve faster decoding while keeping compact model sizes, and we confirm that our method can utilize the efficiency of the double-array structure to achieve a balance between speed and size on translation tasks.

HighlightsWe analyze why semi-supervised backoff language modeling performs poorly.We motivate MAP adaptation of a log-linear language model.We use automatic transcripts as a prior for language model estimation.We show consistent reduction in WER across a range of low-resource conditions. Many under-resourced languages such as Arabic diglossia or Hindi sub-dialects do not have sufficient in-domain text to build strong language models for use with automatic speech recognition (ASR). Semi-supervised language modeling uses a speech-to-text system to produce automatic transcripts from a large amount of in-domain audio typically to augment a small amount of manual transcripts. In contrast to the success of semi-supervised acoustic modeling, conventional language modeling techniques have provided only modest gains. This paper first explains the limitations of back-off language models due to their dependence on long-span n-grams, which are difficult to accurately estimate from automatic transcripts. From this analysis, we motivate a more robust use of the automatic counts as a prior over the estimated parameters of a log-linear language model. We demonstrate consistent gains for semi-supervised language models across a range of low-resource conditions.

Graphical abstractDisplay Omitted HighlightsHuman translations guided language discovery for speech processing.Pronunciation extraction for non-written languages using cross-lingual information.Alignment model Model 3P for cross-lingual word-to-phoneme alignment.Algorithm to deduce phonetic transcriptions of words from Model 3P alignments.Analysis of appropriate source languages based on efficient evaluation measures. In this paper, we study methods to discover words and extract their pronunciations from audio data for non-written and under-resourced languages. We examine the potential and the challenges of pronunciation extraction from phoneme sequences through cross-lingual word-to-phoneme alignment. In our scenario a human translator produces utterances in the (non-written) target language from prompts in a resource-rich source language. We add the resource-rich source language prompts to help the word discovery and pronunciation extraction process. By aligning the source language words to the target language phonemes, we segment the phoneme sequences into word-like chunks. The resulting chunks are interpreted as putative word pronunciations but are very prone to alignment and phoneme recognition errors. Thus we suggest our alignment model Model?3P that is particularly designed for cross-lingual word-to-phoneme alignment. We present two different methods (source word dependent and independent clustering) that extract word pronunciations from word-to-phoneme alignments and compare them. We show that both methods compensate for phoneme recognition and alignment errors. We also extract a parallel corpus consisting of 15 different translations in 10 languages from the Christian Bible to evaluate our alignment model and error recovery methods. For example, based on noisy target language phoneme sequences with 45.1% errors, we build a dictionary for an English Bible with a Spanish Bible translation with 4.5% OOV rate, where 64% of the extracted pronunciations contain no more than one wrong phoneme. Finally, we use the extracted pronunciations in an automatic speech recognition system for the target language and report promising word error rates - given that pronunciation dictionary and language model are learned completely unsupervised and no written form for the target language is required for our approach.

In this paper, we present novel techniques for performing topic adaptation on an n-gram language model. Given training text labeled with topic information, we automatically identify the most relevant topics for new text. We adapt our language model toward these topics using an exponential model, by adjusting probabilities in our model to agree with those found in the topical subset of the training data. For efficiency, we do not normalize the model; that is, we do not require that the "probabilities" in the language model sum to 1. With these techniques, we were able to achieve a modest reduction in speech recognition word-error rate in the Broadcast News domain.

The rise of mobile devices and online learning brings into sharp focus the importance of speech recognition not only for the many languages of the world but also for code-mixed speech, especially where English is the second language. The recognition of code-mixed speech, where the speaker mixes languages within a single utterance, is a challenge for both computers and humans, not least because of the limited training data. We conduct research on a Mandarin-English code-mixed lecture corpus, where Mandarin is the host language and English the guest language, and attempt to find complex features for the recovery of English segments that were misrecognized in the initial recognition pass. We propose a multi-level framework wherein both low-level and high-level cues are jointly considered; we use phonotactic, prosodic, and linguistic cues in addition to acoustic-phonetic cues to discriminate at the frame level between English- and Chinese-language segments. We develop a simple and exact method for CRF feature induction, and improved methods for using cascaded features derived from the training corpus. By additionally tuning the data imbalance ratio between English and Chinese, we demonstrate highly significant improvements over previous work in the recovery of English-language segments, and demonstrate performance superior to DNN-based methods. We demonstrate considerable performance improvements not only with the traditional GMM-HMM recognition paradigm but also with a state-of-the-art hybrid CD-HMM-DNN recognition framework.

Larger n-gram language models (LMs) perform better in statistical machine translation (SMT). However, the existing approaches have two main drawbacks for constructing larger LMs: 1) it is not convenient to obtain larger corpora in the same domain as the bilingual parallel corpora in SMT; 2) most of the previous studies focus on monolingual information from the target corpora only, and redundant n-grams have not been fully utilized in SMT. Nowadays, continuous-space language model (CSLM), especially neural network language model (NNLM), has been shown great improvement in the estimation accuracies of the probabilities for predicting the target words. However, most of these CSLM and NNLM approaches still consider monolingual information only or require additional corpus. In this paper, we propose a novel neural network based bilingual LM growing method. Compared to the existing approaches, the proposed method enables us to use bilingual parallel corpus for LM growing in SMT. The results show that our new method outperforms the existing approaches on both SMT performance and computational efficiency significantly.

In this paper, we propose the use of distance and co-occurrence information of word-pairs to improve language modeling. We have empirically shown that, for history-context sizes of up to ten words, the extracted information about distance and co-occurrence complements the n-gram language model well, for which learning long-history contexts is inherently difficult. Evaluated on the Wall Street Journal and the Switchboard corpora, our proposed model reduces the trigram model perplexity by up to 11.2% and 6.5%, respectively. As compared to the distant bigram model and the trigger model, our proposed model offers a more effective manner of capturing far context information, as verified in terms of perplexity and computational efficiency, i.e., fewer free parameters to be fine-tuned. Experiments using the proposed model for speech recognition, text classification and word prediction tasks showed improved performance.

Machine translation (MT) is defined as a subfield of computational linguistics that investigates the use of computer software to translate text or speech from one natural language to another. It is a key application in the field of natural language processing. At its basic level, MT performs simple substitution of words in one natural language for words in another. Effort to access documents from one language to another leads to the development of machine translation system which involves lots of heterogeneous features and its implementations. Approaches to machine translations are different and each of this approach has its own benefits and drawbacks. This study looks at the various approaches to machine translations and future needs in order to provide more robust and sensible system in the area of natural language processing which will be resistant and impervious to failure regardless of users’ inputs. It is hopeful that researchers in the area of language processing can make use of our valuable improvement and suggestions.

Objectives: The article presents the experiments on the creation of different language models for the Tatar language. N-gram statistical models are used with five different smoothing techniques. Methods: These models can be used in various applications: machine translation systems, spell checking, etc. The study intended to use the patterns in the system of Tatar speech automatic recognition. Taking into account the specifics of the Tatar language, consisting in a rich morphology, speech recognition systems may use not only words but also the building blocks of words as basic modeling units: syllables, morphemes, etc Finding: The following essential elements were chosen for a complete analysis of Tatar language models development: word, morpheme, morph (statistically selected component of a nutshell), the stem and affix chain, syllable and letter. Thus, some models constructed for all combinations of 2-, 3-, 4-grams, smoothing techniques and essential elements of the language. Besides, an experiment showing the possibility of a language model development based on word classes conducted. Conclusion: According to performed experiment results the conclusions are made about the quality of the Tatar language grammar description, the degree of coverage lexicon, and required vocabulary volume for each type of constructed models.

During the last decade, we assist to a major change in the direction that theoretical models used in natural language processing follow. We are moving from rule-based systems to corpus-oriented paradigms. In this paper, we analyze several generative formalisms together with newer statistical and data-oriented linguistic methodologies. We review existing methods belonging to deep or shallow learning applied in various subfields of computational linguistics. The continuous, fast improvements obtained by practical, applied machine learning techniques may lead us to new theoretical developments in the classic models as well. We discuss several scenarios for future approaches.

Speech interfaces to conversational systems have been a focus in academia and industry for over a decade due to its applicability as a natural interface. Speech recognition and speech synthesis constitute the important input and output modules respectively for such spoken interface systems. In this paper, the speech recognition interface for question answering applications is reviewed, and existing limitations are discussed. The existing spoken question answering (QA) systems use an automatic speech recogniser by adapting acoustic and language models for the speech interface and off-the-shelf language processing systems for question interpretation. In the process, the impact of recognition errors and language processing inaccuracies is neglected. It is illustrated in the paper how a semantically rich knowledge graph can be used to solve automatic speech recognition and language processing specific problems. A simple concatenation of a speech recogniser and a natural language processing system is a shallow method for a speech interface. An effort beyond merely concatenating these two units is required to develop a successful spoken question answering system. It is illustrated in this paper how a knowledge graph based structured data can be used to build a unified system combining speech recognition and language understanding. This facilitates the use of a semantically rich data model for speech interface.

This paper describes the architecture of a Machine Translation System with source language as English and target language as Marathi. The basic approach used for the development of this system is Rule Based Machine Translation. The basic algorithm for obtaining the correct word order in the target language was developed based on specific traversals of the parse tree. One of the special features of the system is a Word Sense Disambiguation model. Presently only prepositions will be disambiguated and work is going on for verbs and nouns. The model is a generalized approach based on the categories/domains a word belongs to. Another feature is the target language generation module. The focus is on the grammar structure of the target language that will produce better and smoother translations. The architecture though developed specifically for English – Marathi language pair, may be extended to other language pairs with similar structure. The architecture is partially implemented in the form of Machine Translation system. A lexicon is built for morphological and semantic properties. The results, even at partial implementation stage, are really encouraging.

The n-gram model and its derivatives are both widely applied solutions for Large Vocabulary Continuous Speech Recognition (LVCSR) systems. However, Slavonic languages require a language model that considers word order less strictly than English, i.e. the language that is the subject of most linguistic research. Such a language model is a necessary module in LVCSR systems, because it increases the probability of finding the right word sequences. The aim of the presented work is to create a language module for the Polish language with the application of neural networks. Here, the capabilities of Kohonen's Self-Organized Maps will be explored to find the associations between words in spoken utterances. To fulfill such a task, the application of neural networks to evaluate sequences of words will be presented. Then, the next step of language model development, the network architectures, will be discussed. The network proposed for the construction of the considered model is inspired by the Cocke-Young-Kasami parsing algorithm.

Statistical machine translation is the process of translating languages using machine learning methods. By learning from a large parallel text corpora, Statistical Machine Translation automatically learns how to translate one language into another. This paper will present a method to translate English language to Assamese language using Statistical Machine Translation.

Translation is a changing process from one language to another language. There are various kinds of languages in the world, in most cases, people can communicate through translation. In today's rapid development of the computer, it is natural to consider using a computer to help translate. Machine translation refers to the use of computer translation. Machine translation is a combination of interdisciplinary research in mathematics, linguistics, computer science and other disciplines and use a computer to translate a natural source language into another target language. Machine translation includes a number of key steps which are the process of machine translation plays a very important role.

As the content of a full text page usually focuses on a specific topic, a topic language model adaption method is proposed to improve the recognition performance of homologous offline handwritten Chinese text image. Firstly, the text images are recognized with a character based bi-gram language model. Secondly, the topic of the text image is matched adaptively. Finally, the text image is recognized again with the best matched topic language model. To obtain a tradeoff between the recognition performance and computational complexity, a restricted topic language model adaption method is further presented. The methods have been evaluated on 100 offline Chinese text images. Compared to the general language model, the topic language model adaption has reduced the relative error rate by 11.94%. The restricted topic language model has lessened the running time by 19.22% at the cost of losing 0.35% of the accuracy.

This paper presents an effective approach for unsupervised language model adaptation (LMA) using multiple models in offline recognition of unconstrained handwritten Chinese texts. The domain of the document to recognize is variable and usually unknown a priori, so we use a two-pass recognition strategy with a pre-defined multi-domain language model set. We propose three methods to dynamically generate an adaptive language model to match the text output by first-pass recognition: model selection, model combination and model reconstruction. In model selection, we use the language model with minimum perplexity on the first-pass recognized text. By model combination, we learn the combination weights via minimizing the sum of squared error with both L2-norm and L1-norm regularization. For model reconstruction, we use a group of orthogonal bases to reconstruct a language model with the coefficients learned to match the document to recognize. Moreover, we reduce the storage size of multiple language models using two compression methods of split vector quantization (SVQ) and principal component analysis (PCA). Comprehensive experiments on two public Chinese handwriting databases CASIA-HWDB and HIT-MW show that the proposed unsupervised LMA approach improves the recognition performance impressively, particularly for ancient domain documents with the recognition accuracy improved by 7 percent. Meanwhile, the combination of the two compression methods largely reduces the storage size of language models with little loss of recognition accuracy. HighlightsWe propose an unsupervised language model (LM) adaptation framework for handwritten Chinese text recognition.We use a two-pass recognition strategy with a pre-defined multi-domain LM set.The adaptive LM is dynamically generated by three methods of model selection, model combination and model reconstruction.We compress the LM set by split vector quantization and principal component analysis.

With the popularization of internet, cyber language is used more and more frequently. The network language has formed a unique system of computer language library. So studying of computer language has great significance to the development of the network language. Language library capacity cant be unlimited expansion. To enhance the computer language network function we must use correlation algorithm to simplify the computer language. In this paper, an improved method based on the network language new is proposed. We apply the neural network speech recognition model with multi hierarchy recognition technology to computer language, and put forward TEI model of computer network language identification. We use the VB program to call on the MATLAB software, successfully simplify the complex recognition process of the network language. At the same time, computer network language also has certain reference value for English language learning. It can help learners to avoid the negative effects of network language on the English language.

Pinyin-to-character conversion is an important task for Chinese natural language processing tasks. Previous work mainly focused on n-gram language models and machine learning approaches, or with additional hand-crafted or automatic rule-based post-processing. There are two problems unable to solve for word n-gram language model: out-of-vocabulary word recognition and long-distance grammatical constraints. In this study, we proposed a linear reranking model trying to solve these problems. Our model uses minimum error learning method to combine different sub models, which includes word and character n-gram LMs, part-of-speech tagging model and dependency model. Impact of different sub models on the conversion are fully experimented and analyzed. Results on the Lancaster Corpus of Mandarin Chinese show that our new model outperforms word n-gram language model.

In the last two decades, significant effort has been put into annotating linguistic resources in several languages. Despite this valiant effort, there are still many languages left that have only small amounts of such resources. The goal of this article is to present and investigate a method of propagating information (specifically mentions) from a resource-rich language such as English into a relatively less-resource language such as Arabic. We compare also this approach to its equivalent counterpart using monolingual resources. Part of the investigation is to quantify the contribution of propagating information in different conditions - based on the availability of resources in the target language. Experiments on the language pair Arabic-English show that one can achieve relatively decent performance by propagating information from a language with richer resources such as English into Arabic alone (no resources or models in the source language Arabic). Furthermore, results show that propagated features from English do help improve the Arabic system performance even when used in conjunction with all feature types built from the source language. Experiments also show that using propagated features in conjunction with lexically-derived features only (as can be obtained directly from a mention annotated corpus) brings the system performance at the one obtained in the target language by using feature derived from many linguistic resources, therefore improving the system when such resources are not available.

Speech is the most natural way of human communication and in order to achieve convenient and efficient human-computer interaction implementation of state-of-the-art spoken language technology is necessary. Research in this area has been traditionally focused on several main languages, such as English, French, Spanish, Chinese or Japanese, but some other languages, particularly Eastern European languages, have received much less attention. However, recently, research activities on speech technologies for Czech, Polish, Serbo-Croatian, Russian languages have been steadily increasing. In this paper, we describe our efforts to build an automatic speech recognition (ASR) system for the Russian language with a large vocabulary. Russian is a synthetic and highly inflected language with lots of roots and affixes. This greatly reduces the performance of the ASR systems designed using traditional approaches. In our work, we have taken special attention to the specifics of the Russian language when developing the acoustic, lexical and language models. A special software tool for pronunciation lexicon creation was developed. For the acoustic model, we investigated a combination of knowledge-based and statistical approaches to create several different phoneme sets, the best of which was determined experimentally. For the language model (LM), we introduced a new method that combines syntactical and statistical analysis of the training text data in order to build better n-gram models. Evaluation experiments were performed using two different Russian speech databases and an internally collected text corpus. Among the several phoneme sets we created, the one which achieved the fewest word level recognition errors was the set with 47 phonemes and thus we used it in the following language modeling evaluations. Experiments with 204 thousand words vocabulary ASR were performed to compare the standard statistical n-gram LMs and the language models created using our syntactico-statistical method. The results demonstrated that the proposed language modeling approach is capable of reducing the word recognition errors.

Speech recognition is a classic example of a human/machine interface, typifying many of the difficulties and opportunities of human/machine interaction. In this paper, speech recognition is used as an example of applying turbo processing principles to the general problem of human/machine interface. Speech recognizers frequently involve a model representing phonemic information at a local level, followed by a language model representing information at a nonlocal level. This structure is analogous to the local (e.g., equalizer) and nonlocal (e.g., error correction decoding) elements common in digital communications. Drawing from the analogy of turbo processing for digital communications, turbo speech processing iteratively feeds back the output of the language model to be used as prior probabilities for the phonemic model. This analogy is developed here, and the performance of this turbo model is characterized by using an artificial language model. Using turbo processing, the relative error rate improves significantly, especially in high-noise settings.

Real-time voice-based machine translation is stimulated by many international teams who want to understand each other syntactically as well as semantically. Authors Tiago Duarte, Rafael Prikladnicki, Fabio Calefato, and Filippo Lanubile provide an overview on current technologies for real-time voice-base machine translation.

One key issue in bilingualism is how bilinguals control production, particularly to produce words in the less dominant language. Language switching is one method to investigate control processes. Language switching has been much studied in comprehension, e.g., in lexical decision task, but less so in production. Here we first present a study of language switching in Italian–English adult bilinguals in a naming task for visually presented words. We demonstrate an asymmetric pattern of time costs to switch language, where participants incurred a greater time cost to switch into naming in their dominant language (Italian). In addition, costs were greater where the stimuli were interlingual cognates or homographs than words existing in only one language, implicating lexical competition as a source of the cost. To clarify the operation of control processes, we then present two connectionist models of bilingual naming, based on the previous models of Seidenberg and McClelland (1989), Cohen, Dunbar and McClelland (1990), Gilbert and Shallice (2002), and Karaminis and Thomas (2010). Crucially, both models acquired their differential language dominance via an experience-dependent learning process. The models embody different assumptions about the language control processes that produce the switch cost. We consider which processing assumptions are sufficient to explain asymmetric language switch costs and word class effects on language switching in individual word reading, as well as generating novel predictions for future testing.

Recent research on multilingual statistical machine translation focuses on the usage of pivot languages in order to overcome language resource limitations for certain language pairs. Due to the richness of available language resources, English is, in general, the pivot language of choice. However, factors like language relatedness can also effect the choice of the pivot language for a given language pair, especially for Asian languages, where language resources are currently quite limited. In this article, we provide new insights into what factors make a pivot language effective and investigate the impact of these factors on the overall pivot translation performance for translation between 22 Indo-European and Asian languages. Experimental results using state-of-the-art statistical machine translation techniques revealed that the translation quality of 54.8% of the language pairs improved when a non-English pivot language was chosen. Moreover, 81.0% of system performance variations can be explained by a combination of factors such as language family, vocabulary, sentence length, language perplexity, translation model entropy, reordering, monotonicity, and engine performance.

The aim of works described in this article is to elaborate and experimentally evaluate a consistent method of Language Model (LM) construction for the sake of Polish speech recognition. In the proposed method we tried to take into account the features and specific problems experienced in practical applications of speech recognition in the Polish language, reach inflection, a loose word order and the tendency for short word deletion. The LM is created in five stages. Each successive stage takes the model prepared at the previous stage and modifies or extends it so as to improve its properties. At the first stage, typical methods of LM smoothing are used to create the initial model. Four most frequently used methods of LM construction are here. At the second stage the model is extended in order to take into account words indirectly co-occurring in the corpus. At the next stage, LM modifications are aimed at reduction of short word deletion errors, which occur frequently in Polish speech recognition. The fourth stage extends the model by insertion of words that were not observed in the corpus. Finally the model is modified so as to assure highly accurate recognition of very important utterances. The performance of the methods applied is tested in four language domains.

The language mixing in multi-language speech recognition is one of the hot issues of concern. After analyzing recognition problem, a method to distinguish language with re-class method according to confidence on multi-language recognition result based on Bayesian decision-making rules with minimum error rate and minimum risk was brought out. It can not only avoid cumbersome language recognition in traditional method but also achieve target of decreasing mixing cognition rate. Experiment on Chinese-English mixing recognition shows that the method can distinguish different language and improve speech recognition rate, which has practicality.

Language modeling (LM), providing a principled mechanism to associate quantitative scores to sequences of words or tokens, has long been an interesting yet challenging problem in the field of speech and language processing. The n-gram model is still the predominant method, while a number of disparate LM methods, exploring either lexical co-occurrence or topic cues, have been developed to complement the n-gram model with some success. In this paper, we explore a novel language modeling framework built on top of the notion of relevance for speech recognition, where the relationship between a search history and the word being predicted is discovered through different granularities of semantic context for relevance modeling. Empirical experiments on a large vocabulary continuous speech recognition (LVCSR) task seem to demonstrate that the various language models deduced from our framework are very comparable to existing language models both in terms of perplexity and recognition error rate reductions.

This paper proposes using cross-lingual language modeling with syntactic information for low-resource speech recognition. We propose phrase-level transduction and syntactic reordering for transcribing a resource-poor language and translating it into a resource-rich language, if necessary. The phrase-level transduction is capable of performing n -m cross-lingual transduction. The syntactic reordering serves to model the syntactic discrepancies between the source and target languages. Our purpose is to leverage the statistics in a resource-rich language model to improve the language model of a resource-poor language and at the same time to improve low-resource speech recognition performance. We implement our cross-lingual language model using weighted finite-state transducers (WFSTs), and integrate it into a WFST-based speech recognition search space to output the transcriptions of both resource-poor and resource-rich languages. This creates an integrated speech transcription and translation framework. Evaluations on Cantonese speech transcription and Cantonese to standard Chinese translation tasks show that our proposed approach improves the system performance significantly, with up to 12.5% relative character error rate (CER) reduction over baseline language model interpolation, 6.6% relative CER reduction and 18.5% relative BLEU score improvement, compared to the best word-level transduction approach.

Spoken language recognition refers to the automatic process through which we determine or verify the identity of the language spoken in a speech sample. We study a computational framework that allows such a decision to be made in a quantitative manner. In recent decades, we have made tremendous progress in spoken language recognition, which benefited from technological breakthroughs in related areas, such as signal processing, pattern recognition, cognitive science, and machine learning. In this paper, we attempt to provide an introductory tutorial on the fundamentals of the theory and the state-of-the-art solutions, from both phonological and computational aspects. We also give a comprehensive review of current trends and future research directions using the language recognition evaluation (LRE) formulated by the National Institute of Standards and Technology (NIST) as the case studies.

Conditional random fields (CRFs) are probabilistic sequence models that have been applied in the last decade to a number of applications in audio, speech, and language processing. In this paper, we provide a tutorial overview of CRF technologies, pointing to other resources for more in-depth discussion; in particular, we describe the common linear-chain model as well as a number of common extensions within the CRF family of models. An overview of the mathematical techniques used in training and evaluating these models is also provided, as well as a discussion of the relationships with other probabilistic models. Finally, we survey recent work in speech, audio, and language processing to show how the same CRF technology can be deployed in different scenarios.

After twenty years of disfavor, a technology has returned which imitates the processes of the brain. Natural language experiments (Sejnowski & Rosenberg: 1986) demonstrate that neural network computing architecture can learn from actual spoken language, observe rules of pronunciation, and reproduce sounds from the patterns derived by its own processes. The consequences of neural network computing for natural language processing activities, including second language acquisition and representation, machine translation, and knowledge processing may be more convulsively revolutionary than anything imagined in current technology. This paper introduces neural network concepts to a traditional natural language processing audience.

In spite of all technological advances made in the field of automatic speech recognition (ASR), due to many factors that influence the recognition rate, speech recognition is still a challenging research area. In this paper, the authoress presents a speech recognition system for Romanian language, based on statistical methods, hidden Markov models (HMMs). The system was experimented for a set of voice commands applied to control electronic devices.

It has been argued that recurrent neural network language models are better in capturing long-range dependency than n-gram language models. In this paper, we attempt to verify this claim by investigating the prediction accuracy and the perplexity of these language models as a function of word position, i.e., the position of a word in a sentence. It is expected that as word position increases, the advantage of using recurrent neural network language models over n-gram language models will become more and more evident. On the text corpus of Penn Tree Bank (PTB), a recurrent neural network language model outperforms a trigram language model in both perplexity and word prediction. However, on the AMI meeting corpus, a trigram outperforms a recurrent neural network language model.

Language modeling plays a critical role in many natural language processing (NLP) tasks such as text prediction, machine translation and speech recognition. Traditional statistical language models (e.g. n-gram models) can only offer words that have been seen before and can not capture long word context. Neural language model provides a promising solution to surpass this shortcoming of statistical language model. This paper investigates Recurrent Neural Networks (RNNs) language model for Vietnamese, at character and syllable-levels. Experiments were conducted on a large dataset of 24M syllables, constructed from 1,500 movie subtitles. The experimental results show that our RNN-based language models yield reasonable performance on the movie subtitle dataset. Concretely, our models outperform n-gram language models in term of perplexity score.