| Deyi Xiong (熊德意) |
Neural Machine Translation: Foundations, Principles, Practices and Frontiers (in Chinese, 512 pages)
By Deyi Xiong, Liangyou Li and Meng Zhang
Publisher: Publishing House of Electronics Industry
ISBN: 978-7-121-43752-6
Book Description
This book focuses on neural machine translation (NMT), systmatically introducing the core methods and cutting-edge research topics in neural machine translation. It is divided into four parts with 20 chapters: foundations, principles, practices and frontiers. The Part I (the foundations) presents the history of NMT, the key differences of neural machine translation from statistical machine translation, basic knowledge on neural networks and natural language processing related to NMT. The Part II (the principles) introduces the vanilla NMT models, attention mechanisms of NMT, convolutional NMT and Transformer according to the order of these technologies in which they were invented. The Part III (the practices) elaborates in turn the data preparation, the training of NMT models, the testing and deployment of developed NMT systems in practice. The Part IV (the frontiers) surveys recent work, challenges and open problems in 6 frontier topics, i.e., discourse-level NMT, low-resource and unsupervised NMT, knowledge-enhanced NMT, NMT robustness, multilingual NMT and end-to-end speech-to-text translation & multimodal NMT. In addition, a special feature of this book is that each chapter is accompanied by a short review either on the topic related to the corresponding chapter, or on the history, stories, ideas, philosophies, controversies and norms behind neural machine translation technology, etc.
This book focuses on neural machine translation (NMT), systmatically introducing the core methods and cutting-edge research topics in neural machine translation. It is divided into four parts with 20 chapters: foundations, principles, practices and frontiers. The Part I (the foundations) presents the history of NMT, the key differences of neural machine translation from statistical machine translation, basic knowledge on neural networks and natural language processing related to NMT. The Part II (the principles) introduces the vanilla NMT models, attention mechanisms of NMT, convolutional NMT and Transformer according to the order of these technologies in which they were invented. The Part III (the practices) elaborates in turn the data preparation, the training of NMT models, the testing and deployment of developed NMT systems in practice. The Part IV (the frontiers) surveys recent work, challenges and open problems in 6 frontier topics, i.e., discourse-level NMT, low-resource and unsupervised NMT, knowledge-enhanced NMT, NMT robustness, multilingual NMT and end-to-end speech-to-text translation & multimodal NMT. In addition, a special feature of this book is that each chapter is accompanied by a short review either on the topic related to the corresponding chapter, or on the history, stories, ideas, philosophies, controversies and norms behind neural machine translation technology, etc.
Contents
- Chapter 1: Introduction
Part I: Foundations
- Chapter 2: Neural Networks
- Chapter 3: Natural Language Processing
Part II: Principles
- Chapter 4: Vanilla Neural Machine Translation
- Chapter 5: Attentional NMT
- Chapter 6: Convolutional NMT
- Chapter 7: Transformer
- Chapter 8: Fundamental Problems of NMT and Their Solutions
Part III: Practices
- Chapter 9: Data Preparation
- Chapter 10: Training
- Chapter 11: Testing
- Chapter 12: Deployment
- Chapter 13: System Design and Implementation of NMT
Part IV: Frontiers
- Chapter 14: Discourse-Level NMT
- Chapter 15: Low-Resource and Unsupervised NMT
- Chapter 16: Knowledge-Enhanced NMT
- Chapter 17: Robust NMT
- Chapter 18: Multilingual NMT
- Chapter 19: Multimodal NMT
- Chapter 20: Closing Remarks