Hybrid approaches to machine translation is a book devoted to an extremely interesting and useful topic. The book presents recent research conducted by linguists and practitioners from different multidisciplinary areas in emerging computer science fields.
The goal of the book is to address the “most important developments in machine translation (MT), [which] are achieved [by] combining data-driven and rule-based techniques.” Combinations of these techniques involve different traditional paradigms: the mixture of “linguistic knowledge into statistical approaches to MT, incorporation of data-driven components into rule-based approaches, or statistical and rule-based pre- and post-processing for both types of MT architectures.”
The book consists of one introductory chapter and three parts with seven more chapters. The introductory chapter provides a general hybrid MT overview.
The first part consists of three papers that cover linguistic knowledge necessary for statistical systems. The first chapter, “Controlled Ascent: Imbuing Statistical MT with Linguistic Knowledge,” explores “the intersection of rule-based and statistical approaches in MT with a focus on past and current work at Microsoft Research.” The second chapter, “Hybrid Word Alignment,” proposes a new model that “provides ... informative alignment links, which are offered by both unsupervised and semi-supervised word alignment models.” The last chapter in this part, “Syntax-Based Pre-reordering for Chinese-to-Japanese Statistical Machine Translation,” deals with problems associated with “the translation of language pairs that have different word orders.”
The second part consists of two chapters that “overview how adding machine learning can help MT.” The first chapter in this part, “Machine Learning Applied to Rule-Based Machine Translation,” “describes an approach to resolve morphologically ambiguous verb forms if rule-based decisions are not possible due to parsing or tagging errors.” The second chapter, “Language-Independent Hybrid MT: Comparative Evaluation of Translation Quality,” “reviews the development of a hybrid MT methodology, which is readily portable to new language pairs.”
The third part also consists of two chapters, which report on hybrid natural language processing (NLP) tools. The main focus in these chapters is how to use MT in different tools like dependency parsers, transduction grammars, and word sense disambiguation. The first chapter, “Creating Hybrid Dependency Parsers for Syntax-Based MT,” discusses “how different forms and different hybrid combinations of dependency parses affect the overall output of syntax-based MT both through automatic and manual evaluation.” The last chapter takes care of serious experiments aimed at the MT “of ambiguous lexical items by using WordNet-based unsupervised word sense disambiguation and comparing its results to three MT systems.”
As the chapters are mostly self-contained, the book can be useful for a wide range of readers. It is primarily devoted to MT specialists in wider fields like computational linguistics and machine learning. It is also useful for “translators and managers of translation companies and departments who are interested in recent developments concerning automated translation tools.” It could also be useful for university teachers and students for courses that are devoted to NLP. Some basic knowledge of artificial intelligence, machine learning, and probability theory is expected.
The material is presented in a high-quality manner. Throughout the book, there are many high-quality and illustrative pictures, diagrams, and tables. The book is missing an index, which could have been useful for readers. Finally, at the end of each chapter, the authors list many interesting and up-to-date references for further reading.
