An Investigation of Multilingual TDNN-BLSTM Acoustic Modeling for Hindi Speech Recognition

2021 ◽  
Vol 11 ◽  
Author(s):  
Ankit Kumar ◽  
Rajesh Kumar Aggarwal
Keyword(s):  
Neural Network ◽  
Speech Recognition ◽  
High Accuracy ◽  
Training Data ◽  
Acoustic Modeling ◽  
Training Dataset ◽  
Acoustic Model ◽  
Indian Languages ◽  
Acoustic Models ◽  
Asr System

Background: In India, thousands of languages or dialects are in use. Most Indian dialects are low asset dialects. A well-performing Automatic Speech Recognition (ASR) system for Indian languages is unavailable due to a lack of resources. Hindi is one of them as large vocabulary Hindi speech datasets are not freely available. We have only a few hours of transcribed Hindi speech dataset. There is a lot of time and money involved in creating a well-transcribed speech dataset. Thus, developing a real-time ASR system with a few hours of the training dataset is the most challenging task. The different techniques like data augmentation, semi-supervised training, multilingual architecture, and transfer learning, have been reported in the past to tackle the fewer speech data issues. In this paper, we examine the effect of multilingual acoustic modeling in ASR systems for the Hindi language. Objective: This article’s objective is to develop a high accuracy Hindi ASR system with a reasonable computational load and high accuracy using a few hours of training data. Method: To achieve this goal we used Multilingual training with Time Delay Neural Network- Bidirectional Long Short Term Memory (TDNN-BLSTM) acoustic modeling. Multilingual acoustic modeling has significantly improved the ASR system's performance for low and limited resource languages. The common practice is to train the acoustic model by merging data from similar languages. In this work, we use three Indian languages, namely Hindi, Marathi, and Bengali. Hindi with 2.5 hours of training data and Marathi with 5.5 hours of training data and Bengali with 28.5 hours of transcribed data, was used in this work to train the proposed model. Results: The Kaldi toolkit was used to perform all the experiments. The paper is investigated over three main points. First, we present the monolingual ASR system using various Neural Network (NN) based acoustic models. Second, we show that Recurrent Neural Network (RNN) language modeling helps to improve the ASR performance further. Finally, we show that a multilingual ASR system significantly reduces the Word Error Rate (WER) (absolute 2% WER reduction for Hindi and 3% for the Marathi language). In all the three languages, the proposed TDNN-BLSTM-A multilingual acoustic models help to get the lowest WER. Conclusion: The multilingual hybrid TDNN-BLSTM-A architecture shows a 13.67% relative improvement over the monolingual Hindi ASR system. The best WER of 8.65% was recorded for Hindi ASR. For Marathi and Bengali, the proposed TDNN-BLSTM-A acoustic model reports the best WER of 30.40% and 10.85%.

Acoustic Modeling of Speech Signal using Artificial Neural Network

2016 ◽  
pp. 196-212
Author(s):  
Mousmita Sarma ◽  
Kandarpa Kumar Sarma
Keyword(s):  
Neural Network ◽  
Artificial Neural Network ◽  
Speech Recognition ◽  
Speech Signal ◽  
Acoustic Modeling ◽  
Artificial Neural ◽  
Artificial Neural Network Ann ◽  
Hidden Layer ◽  
Sound Unit ◽  
Asr System

Acoustic modeling of the sound unit is a crucial component of Automatic Speech Recognition (ASR) system. This is the process of establishing statistical representations for the feature vector sequences for a particular sound unit so that a classifier for the entire sound unit used in the ASR system can be designed. Current ASR systems use Hidden Markov Model (HMM) to deal with temporal variability and Gaussian Mixture Model (GMM) for acoustic modeling. Recently machine learning paradigms have been explored for application in speech recognition domain. In this regard, Multi Layer Perception (MLP), Recurrent Neural Network (RNN) etc. are extensively used. Artificial Neural Network (ANN)s are trained by back propagating the error derivatives and therefore have the potential to learn much better models of nonlinear data. Recently, Deep Neural Network (DNN)s with many hidden layer have been up voted by the researchers and have been accepted to be suitable for speech signal modeling. In this chapter various techniques and works on the ANN based acoustic modeling are described.

Acoustic Modeling of Speech Signal using Artificial Neural Network

2015 ◽  
pp. 282-299 ◽  
Author(s):  
Mousmita Sarma ◽  
Kandarpa Kumar Sarma
Keyword(s):  
Neural Network ◽  
Artificial Neural Network ◽  
Speech Recognition ◽  
Speech Signal ◽  
Acoustic Modeling ◽  
Artificial Neural ◽  
Artificial Neural Network Ann ◽  
Hidden Layer ◽  
Sound Unit ◽  
Asr System

Acoustic modeling of the sound unit is a crucial component of Automatic Speech Recognition (ASR) system. This is the process of establishing statistical representations for the feature vector sequences for a particular sound unit so that a classifier for the entire sound unit used in the ASR system can be designed. Current ASR systems use Hidden Markov Model (HMM) to deal with temporal variability and Gaussian Mixture Model (GMM) for acoustic modeling. Recently machine learning paradigms have been explored for application in speech recognition domain. In this regard, Multi Layer Perception (MLP), Recurrent Neural Network (RNN) etc. are extensively used. Artificial Neural Network (ANN)s are trained by back propagating the error derivatives and therefore have the potential to learn much better models of nonlinear data. Recently, Deep Neural Network (DNN)s with many hidden layer have been up voted by the researchers and have been accepted to be suitable for speech signal modeling. In this chapter various techniques and works on the ANN based acoustic modeling are described.

scholarly journals Evaluation of Power Insulator Detection Efficiency with the Use of Limited Training Dataset

2020 ◽  
Vol 10 (6) ◽  
pp. 2104
Author(s):  
Michał Tomaszewski ◽  
Paweł Michalski ◽  
Jakub Osuchowski
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Object Detection ◽  
Convolutional Neural Network ◽  
Deep Neural Networks ◽  
Detection Efficiency ◽  
Training Data ◽  
Training Dataset ◽  
Training Set ◽  
Convolutional Network

This article presents an analysis of the effectiveness of object detection in digital images with the application of a limited quantity of input. The possibility of using a limited set of learning data was achieved by developing a detailed scenario of the task, which strictly defined the conditions of detector operation in the considered case of a convolutional neural network. The described solution utilizes known architectures of deep neural networks in the process of learning and object detection. The article presents comparisons of results from detecting the most popular deep neural networks while maintaining a limited training set composed of a specific number of selected images from diagnostic video. The analyzed input material was recorded during an inspection flight conducted along high-voltage lines. The object detector was built for a power insulator. The main contribution of the presented papier is the evidence that a limited training set (in our case, just 60 training frames) could be used for object detection, assuming an outdoor scenario with low variability of environmental conditions. The decision of which network will generate the best result for such a limited training set is not a trivial task. Conducted research suggests that the deep neural networks will achieve different levels of effectiveness depending on the amount of training data. The most beneficial results were obtained for two convolutional neural networks: the faster region-convolutional neural network (faster R-CNN) and the region-based fully convolutional network (R-FCN). Faster R-CNN reached the highest AP (average precision) at a level of 0.8 for 60 frames. The R-FCN model gained a worse AP result; however, it can be noted that the relationship between the number of input samples and the obtained results has a significantly lower influence than in the case of other CNN models, which, in the authors’ assessment, is a desired feature in the case of a limited training set.

scholarly journals UCSY-SC1: A Myanmar speech corpus for automatic speech recognition

2019 ◽  
Vol 9 (4) ◽  
pp. 3194 ◽  
Author(s):  
Aye Nyein Mon ◽  
Win Pa Pa ◽  
Ye Kyaw Thu
Keyword(s):  
Neural Network ◽  
Speech Recognition ◽  
Automatic Speech Recognition ◽  
Gaussian Mixture ◽  
Error Rates ◽  
Training Data ◽  
Speech Corpus ◽  
Total Size ◽  
Test Sets ◽  
Web News

This paper introduces a speech corpus which is developed for Myanmar Automatic Speech Recognition (ASR) research. Automatic Speech Recognition (ASR) research has been conducted by the researchers around the world to improve their language technologies. Speech corpora are important in developing the ASR and the creation of the corpora is necessary especially for low-resourced languages. Myanmar language can be regarded as a low-resourced language because of lack of pre-created resources for speech processing research. In this work, a speech corpus named UCSY-SC1 (University of Computer Studies Yangon - Speech Corpus1) is created for Myanmar ASR research. The corpus consists of two types of domain: news and daily conversations. The total size of the speech corpus is over 42 hrs. There are 25 hrs of web news and 17 hrs of conversational recorded data.<br />The corpus was collected from 177 females and 84 males for the news data and 42 females and 4 males for conversational domain. This corpus was used as training data for developing Myanmar ASR. Three different types of acoustic models  such as Gaussian Mixture Model (GMM) - Hidden Markov Model (HMM), Deep Neural Network (DNN), and Convolutional Neural Network (CNN) models were built and compared their results. Experiments were conducted on different data  sizes and evaluation is done by two test sets: TestSet1, web news and TestSet2, recorded conversational data. It showed that the performance of Myanmar ASRs using this corpus gave satisfiable results on both test sets. The Myanmar ASR  using this corpus leading to word error rates of 15.61% on TestSet1 and 24.43% on TestSet2.<br /><br />

scholarly journals A De Novo Divide-and-Merge Paradigm for Acoustic Model Optimization in Automatic Speech Recognition

2020 ◽  
Author(s):  
Conghui Tan ◽  
Di Jiang ◽  
Jinhua Peng ◽  
Xueyang Wu ◽  
Qian Xu ◽  
...  
Keyword(s):  
Speech Recognition ◽  
Automatic Speech Recognition ◽  
De Novo ◽  
Superior Performance ◽  
Acoustic Model ◽  
Acoustic Models ◽  
Public Data ◽  
Speech Data ◽  
Low Efficiency ◽  
Novel Algorithms

Due to the rising awareness of privacy protection and the voluminous scale of speech data, it is becoming infeasible for Automatic Speech Recognition (ASR) system developers to train the acoustic model with complete data as before. In this paper, we propose a novel Divide-and-Merge paradigm to solve salient problems plaguing the ASR field. In the Divide phase, multiple acoustic models are trained based upon different subsets of the complete speech data, while in the Merge phase two novel algorithms are utilized to generate a high-quality acoustic model based upon those trained on data subsets. We first propose the Genetic Merge Algorithm (GMA), which is a highly specialized algorithm for optimizing acoustic models but suffers from low efficiency. We further propose the SGD-Based Optimizational Merge Algorithm (SOMA), which effectively alleviates the efficiency bottleneck of GMA and maintains superior performance. Extensive experiments on public data show that the proposed methods can significantly outperform the state-of-the-art.

Combining Self-supervised Learning and Active Learning for Disfluency Detection

2022 ◽  
Vol 21 (3) ◽  
pp. 1-25
Author(s):  
Shaolei Wang ◽  
Zhongyuan Wang ◽  
Wanxiang Che ◽  
Sendong Zhao ◽  
Ting Liu
Keyword(s):  
Neural Network ◽  
Active Learning ◽  
Supervised Learning ◽  
Large Scale ◽  
Training Data ◽  
Fine Tuning ◽  
Training Dataset ◽  
Performance Gap ◽  
Annotation Costs ◽  
Trained Neural Network

Spoken language is fundamentally different from the written language in that it contains frequent disfluencies or parts of an utterance that are corrected by the speaker. Disfluency detection (removing these disfluencies) is desirable to clean the input for use in downstream NLP tasks. Most existing approaches to disfluency detection heavily rely on human-annotated data, which is scarce and expensive to obtain in practice. To tackle the training data bottleneck, in this work, we investigate methods for combining self-supervised learning and active learning for disfluency detection. First, we construct large-scale pseudo training data by randomly adding or deleting words from unlabeled data and propose two self-supervised pre-training tasks: (i) a tagging task to detect the added noisy words and (ii) sentence classification to distinguish original sentences from grammatically incorrect sentences. We then combine these two tasks to jointly pre-train a neural network. The pre-trained neural network is then fine-tuned using human-annotated disfluency detection training data. The self-supervised learning method can capture task-special knowledge for disfluency detection and achieve better performance when fine-tuning on a small annotated dataset compared to other supervised methods. However, limited in that the pseudo training data are generated based on simple heuristics and cannot fully cover all the disfluency patterns, there is still a performance gap compared to the supervised models trained on the full training dataset. We further explore how to bridge the performance gap by integrating active learning during the fine-tuning process. Active learning strives to reduce annotation costs by choosing the most critical examples to label and can address the weakness of self-supervised learning with a small annotated dataset. We show that by combining self-supervised learning with active learning, our model is able to match state-of-the-art performance with just about 10% of the original training data on both the commonly used English Switchboard test set and a set of in-house annotated Chinese data.

Sign in / Sign up

Export Citation Format

Share Document