Large Vocabulary Continuous Speech Recognition Research Articles

Deep convolutional neural networks-based features for Indonesian large vocabulary speech recognition

There are great interests in developing speech recognition using deep learning technologies due to their capability to model the complexity of pronunciations, syntax, and language rules of speech data better than the traditional hidden Markov model (HMM) do. But, the availability of large amount of data is necessary for deep learning-based speech recognition to be effective. While this is not a problem for mainstream languages such as English or Chinese, this is not the case for non-mainstream languages such as Indonesian. To overcome this limitation, we present deep features based on convolutional neural networks (CNN) for Indonesian large vocabulary continuous speech recognition in this paper. The CNN is trained discriminatively which is different from usual deep learning implementations where the networks are trained generatively. Our evaluations show that the proposed method on Indonesian speech data achieves 7.26% and 9.01% error reduction rates over the state-of-the-art deep belief networks-deep neural networks (DBN-DNN) for large vocabulary continuous speech recognition (LVCSR), with Mel frequency cepstral coefficients (MFCC) and filterbank (FBANK) used as features, respectively. An error reduction rate of 6.13% is achieved compared to CNN-DNN with generative training.

Perspectivas sobre la traducción automática del habla

En Japón se han llevado a cabo muchas actividades de investigación acerca de la traducción automática del habla. Este artículo pretende ofrecer una visión general de dichas actividades y presentar las que se han realizado más recientemente. El sistema S2ST está formado básicamente por tres módulos: el reconocimiento automático del habla continua y de amplios vocabularios (Automatic Speech Recognition, ASR), la traducción automática de textos (Machine translation, MT) y la conversión de texto a voz (Text-to-Speech Synthesis, TTS). Todos los módulos deben ser plurilingües, por lo cual se requieren discursos y corpus multilingües para los modelos de formación. El rendimiento del sistema S2ST mejora considerablemente por medio de un aprendizaje profundo y grandes corpus formativos. Sin embargo, todavía hace falta tratar diversos aspectos, com la simultaneidad, la paralingüística, la dependencia del contexto y de la situación, la intención y la dependencia cultural. Por todo ello, repasaremos las actividades de investigación actuales y discutiremos varias cuestiones relacionadas con la traducción automática del habla de última generación.

Reduced Memory Viterbi Decoding for Hardware-accelerated Speech Recognition

Large Vocabulary Continuous Speech Recognition systems require Viterbi searching through a large state space to find the most probable sequence of phonemes that led to a given sound sample. This needs storing and updating of a large Active State List (ASL) in the on-chip memory (OCM) at regular intervals (called frames), which poses a major performance bottleneck for speech decoding. Most works use hash tables for OCM storage while beam-width pruning to restrict the ASL size. To achieve a decent accuracy and performance, a large OCM, numerous acoustic probability computations, and DRAM accesses are incurred. We propose to use a binary search tree for ASL storage and a max heap data structure to track the worst cost state and efficiently replace it when a better state is found. With this approach, the ASL size can be reduced from over 32K to 512 with minimal impact on recognition accuracy for a 7,000-word vocabulary model. This, combined with a caching technique for acoustic scores, reduced the DRAM data accessed by 31 \( \times \) and the acoustic probability computations by 26 \( \times \) . The approach has also been implemented in hardware on a Xilinx Zynq FPGA at 200 MHz using the Vivado SDS compiler. We study the tradeoffs among the amount of OCM used, word error rate, and decoding speed to show the effectiveness of the approach. The resulting implementation is capable of running faster than real time with 91% lesser block-RAMs.

Bangladeshi Bangla speech corpus for automatic speech recognition research

This article reports the development of language resource for Bangladeshi Bangla spoken language (BBSL). Bangladeshi Bangla has inadequate large speech corpora for Large Vocabulary Continuous Speech Recognition (LVCSR) system. The accuracy of the automatic speech recognition (ASR) system rests on the quality of the speech corpus. This work discusses the common issues and activities related to the development of a large speech corpus named ▪ (SUBAK.KO). This corpus is designed to support ASR research in Bangladeshi Bangla. It has been labeled sentence-wise. We have trained this corpus with one of the well-known current End-to-End ASR algorithms, Recurrent Neural Networks (RNNs) with Connectionist Temporal Classification (CTC). To know the strengths and weaknesses, the CER (Character Error Rate) and the WER (Word Error Rate) of the trained RNNCTC model have been observed. Another open-source large Bangla ASR corpus has been trained using the same ASR algorithm. Both trained models have been compared to assess the quality of these corpora. It has been found that SUBAK.KO is a more balanced corpus and considered more regional accented speech variability for a LVCSR system compared to that open-source large Bangla ASR corpus.

Deep learning based large vocabulary continuous speech recognition of an under-resourced language Bangladeshi Bangla

Deep learning based large vocabulary continuous speech recognition of an under-resourced language Bangladeshi Bangla

English Audio Language Retrieval Based on Adaptive Speech-Adjusting Algorithm

For the purpose of language retrieval for English listening, this paper designs and implements a cross-language information retrieval system for English listening. Different implementation methods of cross-language information retrieval, query, and translation are analyzed. The system adopts cross-language information retrieval technology based on bilingual dictionaries. According to the cross-language retrieval system of the existing bilingual dictionaries and monolingual dictionaries, based on the design and implementation of the fuzzy search dictionary lookup mechanism, the existing dictionary lookup mechanism is constructed and analyzed. Aiming at the problem of translation ambiguity in information retrieval systems based on bilingual dictionaries, a disambiguities elimination algorithm based on cooccurrence technology is proposed. In continuous speech, the speed of different speakers in different contexts is very different. Deviation from normal speech speed often leads to recognition errors, which makes recognition performance decline. Considering that the influence of speech speed on the length of speech units increases or decreases synchronously, and there is a strong correlation between the lengths of adjacent speech units, an adaptive speech speed algorithm is proposed based on the framework of implicit Markov model based on the information of the length of speech units. Experiments on number string and large vocabulary continuous speech recognition show that the algorithm is effective.

MACHINE LEARNING FOR SPOKEN LANGUAGE TECHNOLOGY

Spoken language technology is one of the domains in which, in our days, machine learning algorithms and especially neural networks are used. Some applications will pe presented in this paper: detecting overlapped speech on short time frames (till 25ms), emotion recognition from speech (including speech stress detection and deceptive speech detection) and the performances of the last large vocabulary continuous speech recognition systems for Romanian developed in the SpeeD Laboratory, from Research Institute “CAMPUS”, University POLITEHNICA of Bucharest

LET-Decoder: A WFST-Based Lazy-Evaluation Token-Group Decoder With Exact Lattice Generation

We propose a novel lazy-evaluation token-group decoding algorithm with on-the-fly composition of weighted finite-state transducers (WFSTs) for large vocabulary continuous speech recognition. In the standard on-the-fly composition decoder, a base WFST and one or more incremental WFSTs are composed during decoding, and then token passing algorithm is employed to generate the lattice on the composed search space, resulting in substantial computation overhead. To improve speed, the proposed algorithm adopts 1) a token-group method, which groups tokens with the same state in the base WFST on each frame and limits the capacity of the group and 2) a lazy-evaluation method, which does not expand a token group and its source token groups until it processes a word label during decoding. Experiments show that the proposed decoder works notably up to 3 times faster than the standard on-the-fly composition decoder.

Investigation on large vocabulary continuous Kannada speech recognition

Investigation on large vocabulary continuous Kannada speech recognition

Large Vocabulary Continuous Speech Recognition for Nepali Language

Speech Recognition is a widely studied topic for high-resource languages like English and Mandarin. A plethora of publications exist that study the performance of several recognition methods for these languages. However differences in phonetics, accent, language model, etc between any two different languages demand for a study of speech recognition methodologies and components separately for each language. In this paper, we present a comparative study of popular speech recognition methods for Nepali, a low-resource Indo-Aryan language. We describe our approach to building the phonetic dictionary and present our findings for DNN and GMM based techniques with speaker adaptation on 50K vocabulary speech recognition task.

Using Deep Time Delay Neural Network for Slot Filling in Spoken Language Understanding

Modeling the context of a target word is of fundamental importance in predicting the semantic label for slot filling task in Spoken Language Understanding (SLU). Although Recurrent Neural Network (RNN) has shown to successfully achieve the state-of-the-art results for SLU, and Bidirectional RNN is capable of obtaining further improvement by modeling information not only from the past, but also from the future, they only consider limited contextual information of the target word. In order to make the network deeper and hence obtain longer contextual information, we propose to use a multi-layer Time Delay Neural Network (TDNN), which is prevalent in current large vocabulary continuous speech recognition tasks. In particular, we use a TDNN with symmetric time delay offset. To make the stacked TDNN easily trained, residual structures and skip concatenation are adopted. In addition, we further improve the model by introducing ResTDNN-BiLSTM, which combines the advantages of both the residual TDNN and BiLSTM. Experiments on slot filling tasks on the Air Travel Information System (ATIS) and Snips benchmark datasets show the proposed SC-TDNN-C achieves state-of-the-art results without any additional knowledge and data resources. Finally, we review and compare slot filling results by using a variety of existing models and methods.

Subspace Gaussian mixture based language modeling for large vocabulary continuous speech recognition

This paper focuses on adaptable continuous space language modeling approach of combining longer context information of recurrent neural network (RNN) with adaptation ability of subspace Gaussian mixture model (SGMM) which has been widely used in acoustic modeling for automatic speech recognition (ASR).In large vocabulary continuous speech recognition (LVCSR) it is a challenging problem to construct language models that can capture the longer context information of words and ensure generalization and adaptation ability. Recently, language modeling based on RNN and its variants have been broadly studied in this field.The goal of our approach is to obtain the history feature vectors of a word with longer context information and model every word by subspace Gaussian mixture model such as Tandem system used in acoustic modeling for ASR. Also, it is to apply fMLLR adaptation method, which is widely used in SGMM based acoustic modeling, for adaptation of subspace Gaussian mixture based language model (SGMLM).After fMLLR adaptation, SGMLMs based on Top-Down and Bottom-Up obtain WERs of 5.70 % and 6.01%, which are better than 4.15% and 4.61% of that without adaptation, respectively. Also, with fMLLR adaptation, Top-Down and Bottom-Up based SGMLMs yield absolute word error rate reduction of 1.48%, 1.02% and a relative perplexity reduction of 10.02%, 6.46% compared to RNNLM without adaptation, respectively.

A usage of the syllable unit based on morphological statistics in Korean large vocabulary continuous speech recognition system

In large vocabulary continuous speech recognition (LVCSR), it is important in improving the system’s performance to determine reasonably the recognition unit. In Korean continuous speech recognition, a morph rather than a word is used basically as the recognition unit due to Korean’s agglutinative property and a good performance is provided by combining high-frequency morph sequences, which leading to an increase of vocabulary size and high out-of-vocabulary (OOV) rate. Sub-lexical units such as a syllable and a graphone are widely used for inflectional languages, while they have not been introduced successfully for Korean speech recognition, due to a weakness of their linguistic information. In this paper, we investigate a usage of a syllable unit to resolve a mismatch problem between the recognition unit and vocabulary size that have occurred frequently in Korean large vocabulary speech recognition. We apply the local segmentation into syllables based on morphological statistics and perform experiments using the language model (LM) constructed from mixed unit types of morpheme, combined morpheme and syllable. By the proposed model, an absolute reduction of around 0.4% in word error rate (WER) is obtained compared to a traditional LM consisting of morphemes and combined morphemes.

Efficient Search Mechanism from Large Scale Corpora for Domain-Specific Language Modeling in Speech Recognition

With the Internet and the World Wide Web revolution, large corpora in variety of forms are germinating ceaselessly that can be manifested as big data. One obligatory area for the usage of such large corpora is language modeling for large vocabulary continuous speech recognition. Language modeling is an indispensable module in speech recognition architecture, which plays a vital role in reducing the search space during the recognition process. Additionally, the language model that is contiguous to the domain of the speech can dwindle the search space and escalate the recognition accuracy. In this paper, an efficient searching mechanism for domain-specific document retrieval from the large corpora has been elucidated using Elasticsearch which is a distributed and an efficient search engine for big data. This assisted us in tuning the language model in accordance with the domain and also by reducing the search time by more than 90% in comparison to conventional search and retrieval mechanism used in our earlier work. A word level and a phrase level retrieval process for creating domain-specific language model has been implemented. The evaluation of the system is performed on the basis of word error rate (WER) and perplexity (PPL) of the speech recognition system. The results shows nearly 10% decrease on WER and a major reduction in the PPL that helped in boosting the performance of the speech recognition process. From the results, it can be consummated that Elasticsearch is an efficient mechanism for domain specific document retrieval from large corpora rather than using topic modeling toolkits

An Overview of End-to-End Automatic Speech Recognition

Automatic speech recognition, especially large vocabulary continuous speech recognition, is an important issue in the field of machine learning. For a long time, the hidden Markov model (HMM)-Gaussian mixed model (GMM) has been the mainstream speech recognition framework. But recently, HMM-deep neural network (DNN) model and the end-to-end model using deep learning has achieved performance beyond HMM-GMM. Both using deep learning techniques, these two models have comparable performances. However, the HMM-DNN model itself is limited by various unfavorable factors such as data forced segmentation alignment, independent hypothesis, and multi-module individual training inherited from HMM, while the end-to-end model has a simplified model, joint training, direct output, no need to force data alignment and other advantages. Therefore, the end-to-end model is an important research direction of speech recognition. In this paper we review the development of end-to-end model. This paper first introduces the basic ideas, advantages and disadvantages of HMM-based model and end-to-end models, and points out that end-to-end model is the development direction of speech recognition. Then the article focuses on the principles, progress and research hotspots of three different end-to-end models, which are connectionist temporal classification (CTC)-based, recurrent neural network (RNN)-transducer and attention-based, and makes theoretically and experimentally detailed comparisons. Their respective advantages and disadvantages and the possible future development of the end-to-end model are finally pointed out. Automatic speech recognition is a pattern recognition task in the field of computer science, which is a subject area of Symmetry.

Standard Yorùbá context dependent tone identification using Multi-Class Support Vector Machine (MSVM)

Most state-of-the-art large vocabulary continuous speech recognition systems employ context dependent (CD) phone units, however, the CD phone units are not efficient in capturing long-term spectral dependencies of tone in most tone languages. The Standard Yorùbá (SY) is a language composed of syllable with tones and requires different method for the acoustic modeling. In this paper, a context dependent tone acoustic model was developed. Tone unit is assumed as syllables, amplitude magnified difference function (AMDF) was used to derive the utterance wide F contour, followed by automatic syllabification and tri-syllable forced alignment with speech phonetization alignment and syllabification SPPAS tool. For classification of the context dependent (CD) tone, slope and intercept of F values were extracted from each segmented unit. Supervised clustering scheme was utilized to partition CD tri-tone based on category and normalized based on some statistics to derive the acoustic feature vectors. Multi-class support vector machine (MSVM) was used for tri-tone training. From the experimental results, it was observed that the word recognition accuracy obtained from the MSVM tri-tone system based on dynamic programming tone embedded features was comparable with phone features. A best parameter tuning was obtained for 10-fold cross validation and overall accuracy was 97.5678%. In term of word error rate (WER), the MSVM CD tri-tone system outperforms the hidden Markov model tri-phone system with WER of 44.47%.Keywords: Syllabification, Standard Yorùbá, Context Dependent Tone, Tri-tone Recognition

Binary neural networks for speech recognition

Recently, deep neural networks (DNNs) significantly outperform Gaussian mixture models in acoustic modeling for speech recognition. However, the substantial increase in computational load during the inference stage makes deep models difficult to directly deploy on low-power embedded devices. To alleviate this issue, structure sparseness and low precision fixed-point quantization have been applied widely. In this work, binary neural networks for speech recognition are developed to reduce the computational cost during the inference stage. A fast implementation of binary matrix multiplication is introduced. On modern central processing unit (CPU) and graphics processing unit (GPU) architectures, a 5–7 times speedup compared with full precision floatingpoint matrix multiplication can be achieved in real applications. Several kinds of binary neural networks and related model optimization algorithms are developed for large vocabulary continuous speech recognition acoustic modeling. In addition, to improve the accuracy of binary models, knowledge distillation from the normal full precision floating-point model to the compressed binary model is explored. Experiments on the standard Switchboard speech recognition task show that the proposed binary neural networks can deliver 3–4 times speedup over the normal full precision deep models. With the knowledge distillation from the normal floating-point models, the binary DNNs or binary convolutional neural networks (CNNs) can restrict the word error rate (WER) degradation to within 15.0%, compared to the normal full precision floating-point DNNs or CNNs, respectively. Particularly for the binary CNN with binarization only on the convolutional layers, the WER degradation is very small and is almost negligible with the proposed approach.

Training of reduced-rank linear transformations for multi-layer polynomial acoustic features for speech recognition

The use of higher-order polynomial acoustic features can improve the performance of automatic speech recognition (ASR). However, dimensionality of polynomial representation can be prohibitively large, making acoustic model training using polynomial features infeasible for large vocabulary ASR systems. This paper presents a multi-layer polynomial training framework for acoustic modeling, which recursively expands the acoustic features into their second-order polynomial feature space. After each expansion the dimensionality of resultant features is reduced by a linear transformation. Experimental results obtained for two large-vocabulary continuous speech recognition tasks show that the proposed method outperforms conventional mixture models. More recently the acoustic modelling community has shifted its focus to deep neural networks. We also train multi-layer polynomial features in a similar way: allowing backpropagation and using mean-normalized stochastic gradient descent algorithm. This has led to encouraging results. Specifically, appending a sigmoid-based feed-forward deep neural network with a final polynomial layer has resulted in significant word error rate improvement.

Flat-Start Single-Stage Discriminatively Trained HMM-Based Models for ASR

In recent years, end-to-end approaches to automatic speech recognition have received considerable attention as they are much faster in terms of preparing resources. However, conventional multistage approaches, which rely on a pipeline of training hidden Markov models (HMM)-GMM models and tree-building steps still give the state-of-the-art results on most databases. In this study, we investigate flat-start one-stage training of neural networks using lattice-free maximum mutual information (LF-MMI) objective function with HMM for large vocabulary continuous speech recognition. We thoroughly look into different issues that arise in such a setup and propose a standalone system, which achieves word error rates (WER) comparable with that of the state-of-the-art multi-stage systems while being much faster to prepare. We propose to use full biphones to enable flat-start context-dependent (CD) modeling and show through experiments that our CD modeling approach can be almost as effective as regular tree-based CD modeling. We show that our flat-start LF-MMI setup together with this tree-free CD modeling technique achieves 10 to 25 % relative WER reduction compared to other end-to-end methods on well-known databases. The improvements are larger for smaller databases.

Sequence discriminative training for deep learning based acoustic keyword spotting

Speech recognition is a sequence prediction problem. Besides employing various deep learning approaches for frame-level classification, sequence-level discriminative training has been proved to be indispensable to achieve the state-of-the-art performance in large vocabulary continuous speech recognition (LVCSR). However, keyword spotting (KWS), as one of the most common speech recognition tasks, almost only benefits from frame-level deep learning due to the difficulty of getting competing sequence hypotheses. The few studies on sequence discriminative training for KWS are limited for fixed vocabulary or LVCSR based methods and have not been compared to the state-of-the-art deep learning based KWS approaches. In this paper, a sequence discriminative training framework is proposed for both fixed vocabulary and unrestricted acoustic KWS. Sequence discriminative training for both sequence-level generative and discriminative models are systematically investigated. By introducing word-independent phone lattices or non-keyword blank symbols to construct competing hypotheses, feasible and efficient sequence discriminative training approaches are proposed for acoustic KWS. Experiments showed that the proposed approaches obtained consistent and significant improvement in both fixed vocabulary and unrestricted KWS tasks, compared to previous frame-level deep learning based acoustic KWS methods.