Language Identification System Research Articles

Sparse coding of i-vector/JFA latent vector over ensemble dictionaries for language identification systems

Computing sparse representation (SR) over an exemplar dictionary is time consuming and computationally expensive for large dictionary size. This also requires huge memory requirement for saving the dictionary. In order to reduce the latency and to achieve some diversity, ensemble of exemplar dictionary based language identification (LID) system is explored. The full diversity can be obtained if each of the exemplar dictionary contains only one feature vector from each of the language class. To achieve full diversity, a large number of multiple dictionaries are required; thus needs to compute SR for a particular test utterance as many times. The other solution to reduce the latency is to use a learned dictionary. The dictionary may contain unequal number of dictionary atoms and it is not guaranteed that each language class information is present. It totally depends upon the number of data and its variations. Motivated by this, language specific dictionary is learned, and then concatenated to form a single learned dictionary. Furthermore, to overcome the problem of ensemble exemplar dictionary based LID system, we investigated the ensemble of learned-exemplar dictionary based LID system. The proposed approach achieves the same diversity and latency as that of ensemble exemplar dictionary with reduced number of learned dictionaries. The proposed techniques are applied on two spoken utterance representations: the i-vector and the JFA latent vector. The experiments are performed on 2007 NIST LRE, 2009 NIST LRE and AP17-OLR datasets in closed set condition.

Language and Gender Classification of Speech Files Using Supervised Machine Learning Methods

ABSTRACTMany language identification (LID) systems are based on language models using techniques that consider the fluctuation of speech over time. Considering these fluctuations necessitates longer recording intervals to obtain reasonable accuracy. Our research extracts features from short recording intervals to enable successful classification of spoken language. The feature extraction process is based on frames of 20 ms, whereas most previous LIDs presented results based on much longer frames (3 s or longer). We defined and implemented 200 features divided into four feature sets: cepstrum features, RASTA features, spectrum features, and waveform features. We applied eight machine learning (ML) methods on the features that were extracted from a corpus containing speech files in 10 languages from the Oregon Graduate Institute (OGI) telephone speech database and compared their performances using extensive experimental evaluation. The best optimized classification results were achieved by random forest (RF): from 76.29% on 10 languages to 89.18% on 2 languages. These results are better or comparable to the state-of-the-art results for the OGI database. Another set of experiments that was performed was gender classification from 2 to 10 languages. The accuracy and the F measure values for the RF method for all the language experiments were greater than or equal to 90.05%.

Smoothed n-gram based models for tweet language identification: A case study of the Brazilian and European Portuguese national varieties

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.

Automatic Language Identification

Automatic Language Identification (LID) is the process of automatically identifying the language of spoken utterance or written material. LID has received much attention due to its application to major areas of research and long-aspired dreams in computational sciences, namely Machine Translation (MT), Speech Recognition (SR) and Data Mining (DM). A considerable increase in the amount of and access to data provided not only by experts but also by users all over the Internet has resulted into both the development of different approaches in the area of LID – so as to generate more efficient systems – as well as major challenges that are still in the eye of the storm of this field. Despite the fact that the current approaches have accomplished considerable success, future research concerning some issues remains on the table. The aim of this paper shall not be to describe the historic background of this field of studies, but rather to provide an overview of the current state of LID systems, as well as to classify the approaches developed to accomplish them. LID systems have advanced and are continuously evolving. Some of the issues that need special attention and improvement are semantics, the identification of various dialects and varieties of a language, identification of spelling errors, data retrieval, multilingual documents, MT and speech-to-speech translation. Methods applied to date have been good from a technical point of view, but not from a semantic one.

Performance Enhancement of Indian LID System Using Spectral Processing

Language Identification is a system of recognizing the dialect from a speech sample articulated by an obscure speaker. In this paper, we tend to develop an efficient baseline system for language identification by Mel Frequency Cepstral Coefficients (MFCC) and Gaussian Mixture Model (GMM). The Language Identification (LID) performance is analyzed in speaker independent approach. LID system is developed using IITKGP-MLILSC database. To enhance the LID performance under background noise we use some enhancement techniques like Spectral Subtraction (SS) and Minimum Mean Square Error (MMSE). Here, we present the performance of LID system by varying the amount of training data, the length of testing samples and the background noise.

Parametric representation of excitation source information for language identification

In this work, the linear prediction (LP) residual signal has been parameterized to capture the excitation source information for language identification (LID) study. LP residual signal has been processed at three different levels: sub-segmental, segmental and supra-segmental levels to demonstrate different aspects of language-specific excitation source information. Proposed excitation source features have been evaluated on 27 Indian languages from Indian Institute of Technology Kharagpur-Multi Lingual Indian Language Speech Corpus (IITKGP-MLILSC), Oregon Graduate Institute Multi-Language Telephone-based Speech (OGI-MLTS) and National Institute of Standards and Technology Language Recognition Evaluation (NIST LRE) 2011 corpora. LID systems were developed using Gaussian mixture model (GMM) and i-vector based approaches. Experimental results have shown that segmental level parametric features provide better identification accuracy (62%), compared to sub-segmental (40%) and supra-segmental level (34%) features. Excitation source features obtained from three levels show distinct language-specific evidence. Therefore, the scores from all three levels are combined to obtain the complete excitation source information for the LID task. LID performances achieved from both the excitation source and vocal tract system are compared. Finally, the scores obtained by processing the vocal tract and excitation source features are combined to achieve better improvement in LID accuracy. The best recognition accuracies obtained from stage-IV integrated LID systems I, II and III are 69%, 70% and 72% respectively.

Two approaches-based L2-SVMs reduced to MEB problems for dialect identification

The recent progress in speech and vision has issued from the increased use of machine learning. Not only does the machine learning provides many useful tools, it also help us to understand existing algorithms and their connections in a new light. As a powerful tool in machine learning, support vector machine SVM leads to an expensive computational cost in the training phase due to the large number of original training samples, while minimal enclosing ball MEB presents limitations dealing with a large dataset. The training computation increases as data size becomes large, hence in this paper, we propose two improved approaches that handle this problem in huge dataset used in different domains. These approaches, based on L2-SVMs reduced to MEB problems result in a reduced data optimally matched to the input demands of different background of systems such as Universal Background Model architectures in language recognition and identification systems. We experiment on speech information based on acoustic shifted delta coefficient feature vectors applied in GMM-based dialect identification system where all data outer the ball defined by MEB are eliminated and the training time is reduced. Further numerical experiments on some real-world datasets show proof of the usefulness of our approaches in the field of data mining.

Combining the evidences of temporal and spectral enhancement techniques for improving the performance of Indian language identification system in the presence of background noise

Language Identification has gained significant importance in recent years, both in research and commercial market place, demanding an improvement in the ability of machines to distinguish between l...

TweetLID: a benchmark for tweet language identification

Language identification, as the task of determining the language a given text is written in, has progressed substantially in recent decades. However, three main issues remain still unresolved: (1) distinction of similar languages, (2) detection of multilingualism in a single document, and (3) identifying the language of short texts. In this paper, we describe our work on the development of a benchmark to encourage further research in these three directions, set forth an evaluation framework suitable for the task, and make a dataset of annotated tweets publicly available for research purposes. We also describe the shared task we organized to validate and assess the evaluation framework and dataset with systems submitted by seven different participants, and analyze the performance of these systems. The evaluation of the results submitted by the participants of the shared task helped us shed some light on the shortcomings of state-of-the-art language identification systems, and gives insight into the extent to which the brevity, multilingualism, and language similarity found in texts exacerbate the performance of language identifiers. Our dataset with nearly 35,000 tweets and the evaluation framework provide researchers and practitioners with suitable resources to further study the aforementioned issues on language identification within a common setting that enables to compare results with one another.

Implicit excitation source features for robust language identification

In present work, the robustness of excitation source features has been analyzed for language identification (LID) task. The raw samples of linear prediction (LP) residual signal, its magnitude and phase components are processed at sub-segmental, segmental and supra-segmental levels for capturing the robust language-specific phonotactic information. Present LID study has been carried out on 27 Indian languages from Indian Institute of Technology Kharagpur-Multi Lingual Indian Language Speech Corpus (IITKGP-MLILSC). Gaussian mixture models are used to develop the LID systems using robust language-specific excitation source information. Robustness of excitation source information has been evinced in view of (i) background noise, (ii) varying amount of training data and (iii) varying length of test samples. Finally, the robustness of proposed excitation source features is compared with the well-known spectral features using LID performances obtained from IITKGP-MLILSC database. Segmental level excitation source features obtained from raw samples of LP residual signal and its phase component perform better at low SNR levels, compared with the vocal tract features.

Significance of GMM-UBM based Modelling for Indian Language Identification

Abstract Most of the Indian languages are originated from Devanagari, the script of the Sanskrit language. In-spite of similarity in phoneme sets, every language its own influence on the phonotactic constraints of speech in that language. A modelling technique that is capable of capturing the slightest variations imparted by the language is a pre-requisite for developing a language identification system (LID). Use of Gaussian mixture modelling technique with a large number of mixture components demands a large training data for each language class, which is hard to collect and handle. In this work, phonotactic variations imparted by the different languages are modelled using Gaussian mixture modelling with a universal background model (GMM-UBM) technique. In GMM-UBM based modelling certain amount of data from all the language classes is pooled to develop a universal background model (UBM) and the model is adapted to each class. Spectral features (MFCC) are employed to represent the language specific phonotactic information of speech in different languages. During the present study, LID systems are developed using the speech samples from IITKGP-MLILSC. In this work, performance of the proposed GMM-UBM based LID system is compared with conventional GMM based LID system. An average improvement of 7–8% is observed due to the use of UBM-based modelling of developing a LID system.

GMM based Language Identification using MFCC and SDC Features

Language Identification (LID) is one of the most popular areas of research in speech signal processing. Now a day's lots of approaches have been used to improve performance of LID system which includes Parallel Phone Recognition Language Modeling (PPRLM), Support Vector Machine (SVM) and general Gaussian Mixture Model (GMM) etc. The state-of-art LID system has been utilised lots of feature vectors like LPCC, MFCC, SDC and prosodic. Although fusion of prosodic features with MFCC features shows some improvement in the performance of the LID system. But still it is not sufficient. In this paper, a baseline system for the LID system in multilingual environments has been developed using GMM as a classifier and MFCC combined with Shifted-Delta- Cepstral (SDC) as front end processing feature vectors. In this works, we used the Arunachali Language Speech Database (ALS-DB), a multilingual and multichannel speech corpus which was recently collected from the four local languages namely Adi, Apatani, Galo and Nyishi in Arunachal Pradesh including Hindi and English as secondary languages.The performance of the LID system has been improved by combing MFCC and SDC features than its individual performances. The minimum ERR rates for the features MFCC and SDC individually are 19.70% and 11.83% respectively while minimum ERR rate for the combined features both MFCC and SDC is 6.40%.Approximately 15.00% and 6.00% of performance of the LID system has been improved while using the combining features of MFCC with SDC over the baseline systems that using MFCC and SDC features in individual respectively.

GMM based language identification system using robust features

In this work, we have proposed new feature vectors for spoken language identification (LID) system. The Mel frequency cepstral coefficients (MFCC) and formant frequencies derived using short-time window speech signal. Formant frequencies are extracted from linear prediction (LP) analysis of speech signal. Using these two kind of features of speech signal, new feature vectors are derived using cluster based computation. A GMM based classifier has been designed using these new feature vectors. The language specific apriori knowledge is applied on the recognition output. The experiments are carried out on OGI database and LID recognition performance is improved.

An overview of the development of resources, techniques, and, systems for Indian spoken languages

India possesses a large variety of languages and dialects spoken in different parts of the country. These languages possess some unique linguistic, phonological, and phonetic properties different from European languages. Research is being done in several of Indian languages—such as Hindi, Bangla, etc. to study the articulatory, acoustic—phonetic and prosodic nature for the purpose of creating standards of phonetic representation of phonemes and Pronunciation Lexicon in Indian Languages. Comprehensive and task specific language corpora, speech databases in laboratory as well as in mobile communication situation and the tools/technics required for processing of speech signals are being developed. The emphasis is on developing multilingual human-machine interaction systems. Some of the recently developed systems include multi-lingual speech recognition system for voice enabled services, multilingual text to speech synthesis system, speaker and language identification system for general purpose and forensic applications. Recognition of emotions in spoken speech, spoken language translation system, etc. The paper presents an overview of such studies conducted in various laboratories, academic institutions, and industries in India pertaining to these areas. The technologies used for data collection, processing, and recognition/ synthesis, etc., utilized and status of the development have been mentioned.

Novel Long-Term Information Based Language Identification

A novel long-term information feature for language identification called shifted cepstra curve (SCC) is presented in this paper. Long-term information consists of information over multiple frames, which are commonly used in language identification systems. For instance, in parallel phone recognition language model (PPRLM), the feature vector contains not only information surpassing multiple frames but also linguistic knowledge [1]. However high computational cost for modeling linguistic gram may preclude their use in tasks which demand low memory. By contrast, experiments have proved that linguistics independent long-term information can also achieve high performance in language identification with much lower computing cost. For instance, Shifted delta cepstra (SDC) is employed as long-term information features to establish GMM based language identification systems, which can achieve comparative or even superior performance to PPRLM [2-4]. While, though SDC feature could model the dynamical information across multiple frames, the approximation is not precise enough for language identification. Thus, this paper further presents a new method for linguistics independent long-term information extraction by curve fitting algorithm. Experiments show that new feature based language identification systems are superior to not only short-term features based but also SDC features based systems.

Comparison of VQ and GMM approach for identifying Indian languages

Language identification (LID) is the process of converting an acoustic signal captured by microphone or telephone into a set of words of a particular language in real time thus controlling the computer by the use of spoken commands. The software for LID generally requires an initial training using appropriate classification algorithms in order to teach the software to recognise the language spoken by the user. The LID system we have designed here comprises of four Indian spoken languages-Assamese, Bengali, English and Hindi. Experiments were carried out based on our own recorded standard database consisting of 50 speakers where speech features were extracted using MFCCs. After that VQ and GMM statistical techniques were applied for classification of our designed system followed by testing. Results show that for the vector quantisation approach, accuracy increases almost uniformly with increase in codebook size. Accuracy is maximum at higher codebook sizes for the VQ approach. Comparison of VQ and GMM approach at higher mixture order shows that VQ approach is comparatively better than GMM approach for all the four languages.

A GMM-BASED HIERARCHICAL AUTOMATIC LANGUAGE IDENTIFICATION SYSTEM FOR INDIAN LANGUAGES

Automatic spoken language identification (LID) is the task of identifying a language from a short utterance of speech by an unknown speaker. This article describes a novel two-level identification system for Indian languages using acoustic features. In the first level, the system identifies the family of the spoken language; the second level aims at identifying the particular language within the corresponding family. The proposed system has been modeled using Gaussian mixture model (GMM) and utilizes the following acoustic features: mel frequency cepstral coefficients (MFCC) and shifted delta cepstrum (SDC). A new database has been created for nine Indian languages. It is shown that a GMM-based LID system using MFCC with delta and acceleration coefficients is performing well, with 80.56% accuracy. The performance accuracy of the GMM-based LID system with SDC is also considerable.

A hierarchical language identification system for Indian languages

Automatic spoken Language IDentification (LID) is the task of identifying the language from a short duration of speech signal uttered by an unknown speaker. In this work, an attempt has been made to develop a two level language identification system for Indian languages using acoustic features. In the first level, the system identifies the family of the spoken language, and then it is fed to the second level which aims at identifying the particular language in the corresponding family. The performance of the system is analyzed for various acoustic features and different classifiers. The suitable acoustic feature and the pattern classification model are suggested for effective identification of Indian languages. The system has been modeled using hidden Markov model (HMM), Gaussian mixture model (GMM) and artificial neural networks (ANN). We studied the discriminative power of the system for the features mel frequency cepstral coefficients (MFCC), MFCC with delta and acceleration coefficients and shifted delta cepstral (SDC) coefficients. Then the LID performance as a function of the different training and testing set sizes has been studied. To carry out the experiments, a new database has been created for 9 Indian languages. It is shown that GMM based LID system using MFCC with delta and acceleration coefficients is performing well with 80.56% accuracy. The performance of GMM based LID system with SDC is also considerable.

Human and computer recognition of regional accents and ethnic groups from British English speech

The paralinguistic information in a speech signal includes clues to the geographical and social background of the speaker. This paper is concerned with automatic extraction of this information from a short segment of speech. A state-of-the-art language identification (LID) system is applied to the problems of regional accent recognition for British English, and ethnic group recognition within a particular accent. We compare the results with human performance and, for accent recognition, the ‘text dependent’ ACCDIST accent recognition measure. For the 14 regional accents of British English in the ABI-1 corpus (good quality read speech), our LID system achieves a recognition accuracy of 89.6%, compared with 95.18% for our best ACCDIST-based system and 58.24% for human listeners. The “Voices across Birmingham” corpus contains significant amounts of telephone conversational speech for the two largest ethnic groups in the city of Birmingham (UK), namely the ‘Asian’ and ‘White’ communities. Our LID system distinguishes between these two groups with an accuracy of 96.51% compared with 90.24% for human listeners. Although direct comparison is difficult, it seems that our LID system performs much better on the standard 12 class NIST 2003 Language Recognition Evaluation task or the two class ethnic group recognition task than on the 14 class regional accent recognition task. We conclude that automatic accent recognition is a challenging task for speech technology, and speculate that the use of natural conversational speech may be advantageous for these types of paralinguistic task.

Identification of Language using Mel-Frequency Cepstral Coefficients (MFCC)

This paper focuses on the task of identifying a language from speech signal. In this paper, we have use Mel-frequency cepstral coefficient as features. Language identification models are developed for fifteen Indian languages namely Assamese, Bangla, Guajarati, Hindi, Kannada, Kashmiri, Malayalam, Marathi, Nepali, Oriya, Punjabi, Rajasthani, Tamil, Telugu and Urdu using these spectral features. The identification of above mentioned languages is carried out using Gaussian mixture model. A Semi natural read database is used for obtaining the language specific information. MFCC is obtained by using linear cosine transform of log power spectrum on a nonlinear mel-frequency scale. This paper shows that the performance of Language identification system is better when trained and tested with twenty nine features as compared to six, eight, thirteen, nineteen and twenty one MFCC features. It means more the number of features we use better the result we get. The average language recognition rate over fifteen Indian languages is around 88\\%.