Deep Learning Approach for Natural Language Processing Speech and Computer Vision Techniques and Use Cases 1st Edition by Ashok Kumar, Karthika Renuka 1000875601 9781000875607

Chapter 1 Introduction

Learning Outcomes

1.1 Introduction

1.1.1 Subsets of Artificial Intelligence

1.1.2 Three Horizons of Deep Learning Applications

1.1.3 Natural Language Processing

1.1.4 Speech Recognition

1.1.5 Computer Vision

1.2 Machine Learning Methods for NLP, Computer Vision (CV), and Speech

1.2.1 Support Vector Machine (SVM)

1.2.2 Bagging

1.2.3 Gradient-boosted Decision Trees (GBDTs)

1.2.4 Naïve Bayes

1.2.5 Logistic Regression

1.2.6 Dimensionality Reduction Techniques

1.3 Tools, Libraries, Datasets, and Resources for the Practitioners

1.3.1 TensorFlow

1.3.2 Keras

1.3.3 Deeplearning4j

1.3.4 Caffe

1.3.5 ONNX

1.3.6 PyTorch

1.3.7 scikit-learn

1.3.8 NumPy

1.3.9 Pandas

1.3.10 NLTK

1.3.11 Gensim

1.3.12 Datasets

1.4 Summary

Bibliography

Chapter 2 Natural Language Processing

Learning Outcomes

2.1 Natural Language Processing

2.2 Generic NLP Pipeline

2.2.1 Data Acquisition

2.2.2 Text Cleaning

2.3 Text Pre-processing

2.3.1 Noise Removal

2.3.2 Stemming

2.3.3 Tokenization

2.3.4 Lemmatization

2.3.5 Stop Word Removal

2.3.6 Parts of Speech Tagging

2.4 Feature Engineering

2.5 Modeling

2.5.1 Start with Simple Heuristics

2.5.2 Building Your Model

2.5.3 Metrics to Build Model

2.6 Evaluation

2.7 Deployment

2.8 Monitoring and Model Updating

2.9 Vector Representation for NLP

2.9.1 One Hot Vector Encoding

2.9.2 Word Embeddings

2.9.3 Bag of Words

2.9.4 TF-IDF

2.9.5 N-gram

2.9.6 Word2Vec

2.9.7 Glove

2.9.8 ElMo

2.10 Language Modeling with n-grams

2.10.1 Evaluating Language Models

2.10.2 Smoothing

2.10.3 Kneser-Ney Smoothing

2.11 Vector Semantics and Embeddings

2.11.1 Lexical Semantics

2.11.2 Vector Semantics

2.11.3 Cosine for Measuring Similarity

2.11.4 Bias and Embeddings

2.12 Summary

Bibliography

Chapter 3 State-of-the-Art Natural Language Processing

Learning Outcomes

3.1 Introduction

3.2 Sequence-to-Sequence Models

3.2.1 Sequence

3.2.2 Sequence Labeling

3.2.3 Sequence Modeling

3.3 Recurrent Neural Networks

3.3.1 Unrolling RNN

3.3.2 RNN-based POS Tagging Use Case

3.3.3 Challenges in RNN

3.4 Attention Mechanisms

3.4.1 Self-attention Mechanism

3.4.2 Multi-head Attention Mechanism

3.4.3 Bahdanau Attention

3.4.4 Luong Attention

3.4.5 Global Attention versus Local Attention

3.4.6 Hierarchical Attention

3.5 Transformer Model

3.5.1 Bidirectional Encoder, Representations, and Transformers (BERT)

3.5.2 GPT3

3.6 Summary

Bibliography

Chapter 4 Applications of Natural Language Processing

Learning Outcomes

4.1 Introduction

4.2 Word Sense Disambiguation

4.2.1 Word Senses

4.2.2 WordNet: A Database of Lexical Relations

4.2.3 Approaches to Word Sense Disambiguation

4.2.4 Applications of Word Sense Disambiguation

4.3 Text Classification

4.3.1 Building the Text Classification Model

4.3.2 Applications of Text Classification

4.3.3 Other Applications

4.4 Sentiment Analysis

4.4.1 Types of Sentiment Analysis

4.5 Spam Email Classification

4.5.1 History of Spam

4.5.2 Spamming Techniques

4.5.3 Types of Spams

4.6 Question Answering

4.6.1 Components of Question Answering System

4.6.2 Information Retrieval-based Factoid Question and Answering

4.6.3 Entity Linking

4.6.4 Knowledge-based Question Answering

4.7 Chatbots and Dialog Systems

4.7.1 Properties of Human Conversation

4.7.2 Chatbots

4.7.3 The Dialog-state Architecture

4.8 Summary

Bibliography

Chapter 5 Fundamentals of Speech Recognition

Learning Outcomes

5.1 Introduction

5.2 Structure of Speech

5.3 Basic Audio Features

5.3.1 Pitch

5.3.2 Timbral Features

5.3.3 Rhythmic Features

5.3.4 MPEG-7 Features

5.4 Characteristics of Speech Recognition System

5.4.1 Pronunciations

5.4.2 Vocabulary

5.4.3 Grammars

5.4.4 Speaker Dependence

5.5 The Working of a Speech Recognition System

5.5.1 Input Speech

5.5.2 Audio Pre-processing

5.5.3 Feature Extraction

5.6 Audio Feature Extraction Techniques

5.6.1 Spectrogram

5.6.2 MFCC

5.6.3 Short-Time Fourier Transform

5.6.4 Linear Prediction Coefficients (LPCC)

5.6.5 Discrete Wavelet Transform (DWT)

5.6.6 Perceptual Linear Prediction (PLP)

5.7 Statistical Speech Recognition

5.7.1 Acoustic Model

5.7.2 Pronunciation Model

5.7.3 Language Model

5.7.4 Conventional ASR Approaches

5.8 Speech Recognition Applications

5.8.1 In Banking

5.8.2 In-Car Systems

5.8.3 Health Care

5.8.4 Experiments by Different Speech Groups for Large-Vocabulary Speech Recognition

5.8.5 Measure of Performance

5.9 Challenges in Speech Recognition

5.9.1 Vocabulary Size

5.9.2 Speaker-Dependent or -Independent

5.9.3 Isolated, Discontinuous, and Continuous Speech

5.9.4 Phonetics

5.9.5 Adverse Conditions

5.10 Open-source Toolkits for Speech Recognition

5.10.1 Frameworks

5.10.2 Additional Tools and Libraries

5.11 Summary

Bibliography

Chapter 6 Deep Learning Models for Speech Recognition

Learning Outcomes

6.1 Traditional Methods of Speech Recognition

6.1.1 Hidden Markov Models (HMMs)

6.1.2 Gaussian Mixture Models (GMMs)

6.1.3 Artificial Neural Network (ANN)

6.1.4 HMM and ANN Acoustic Modeling

6.1.5 Deep Belief Neural Network (DBNN) for Acoustic Modelling

6.2 RNN-based Encoder–Decoder Architecture

6.3 Encoder

6.4 Decoder

6.5 Attention-based Encoder–Decoder Architecture

6.6 Challenges in Traditional ASR and the Motivation for End-to-End ASR

6.7 Summary

Bibliography

Chapter 7 End-to-End Speech Recognition Models

Learning Outcomes

7.1 End-to-End Speech Recognition Models

7.1.1 Definition of End-to-End ASR System

7.1.2 Connectionist Temporal Classification (CTC)

7.1.3 Deep Speech

7.1.4 Deep Speech 2

7.1.5 Listen, Attend, Spell (LAS) Model

7.1.6 JASPER

7.1.7 QuartzNet

7.2 Self-supervised Models for Automatic Speech Recognition

7.2.1 Wav2Vec

7.2.2 Data2Vec

7.2.3 HuBERT

7.3 Online/Streaming ASR

7.3.1 RNN-transducer-Based Streaming ASR

7.3.2 Wav2Letter for Streaming ASR

7.3.3 Conformer Model

7.4 Summary

Bibliography

Chapter 8 Computer Vision Basics

Learning Outcomes

8.1 Introduction

8.1.1 Fundamental Steps for Computer Vision

8.1.2 Fundamental Steps in Digital Image Processing

8.2 Image Segmentation

8.2.1 Steps in Image Segmentation

8.3 Feature Extraction

8.4 Image Classification

8.4.1 Image Classification Using Convolutional Neural Network (CNN)

8.4.2 Convolution Layer

8.4.3 Pooling or Down Sampling Layer

8.4.4 Flattening Layer

8.4.5 Fully Connected Layer

8.4.6 Activation Function

8.5 Tools and Libraries for Computer Vision

8.5.1 OpenCV

8.5.2 MATLAB

8.6 Applications of Computer Vision

8.6.1 Object Detection

8.6.2 Face Recognition

8.6.3 Number Plate Identification

8.6.4 Image-based Search

8.6.5 Medical Imaging

8.7 Summary

Bibliography

Chapter 9 Deep Learning Models for Computer Vision

Learning Outcomes

9.1 Deep Learning for Computer Vision

9.2 Pre-trained Architectures for Computer Vision

9.2.1 LeNet

9.2.2 AlexNet

9.2.3 VGG

9.2.4 Inception

9.2.5 R-CNN

9.2.6 Fast R-CNN

9.2.7 Faster R-CNN

9.2.8 Mask R-CNN

9.2.9 YOLO

9.3 Summary

Bibliography

Chapter 10 Applications of Computer Vision

Learning Outcomes

10.1 Introduction

10.2 Optical Character Recognition

10.2.1 Code Snippets

10.2.2 Result Analysis

10.3 Face and Facial Expression Recognition

10.3.1 Face Recognition

10.3.2 Facial Recognition System

10.3.3 Major Challenges in Recognizing Face Expression

10.3.4 Result Analysis

10.4 Visual-based Gesture Recognition

10.4.1 Framework Used

10.4.2 Code Snippets

10.4.3 Result Analysis

10.4.4 Major Challenges in Gesture Recognition

10.5 Posture Detection and Correction

10.5.1 Framework Used

10.5.2 Squats

10.5.3 Result Analysis

10.6 Summary

Bibliography

Index