Machine Translation Review
No. 13, December 2002
ISSN: 1358-8346
This page URL: http://www.bcs.org.uk/siggroup/nalatran/mtreview/mtr-13/7.htm
Book Review
Translation and Information Technology
(2002), The Chinese University of Hong Kong: Chinese University Press
ISBN 962-996-077-X, Price US$18
by Chan Sin-wai (ed.)
Introducing this book, the editor, who heads the Department of Translation at the Chinese University of Hong Kong, draws attention to the fact that MT systems designed by computational linguists and focusing on formalistic analyses of language have so far produced rather disappointing results. He argues for a greater reference to ‘language in use’ rather than a continuing emphasis on grammatical models and suggests that practising translators will in future become more closely involved in MT development. Whether the collection of thirteen papers that follows bears out these points is for the reader to decide, but it covers a fascinating range of systems and work in progress in the field of MT. Particular attention is paid to systems involving Chinese and under development in China.
The volume is divided into three thematic sections: methodology and application, terminology, and critique and training.
The first section describes two Chinese-English MT systems. In the first of these Chan Sin-wai presents TransRecipe, an MT program which combines corpus-based, example-based, pattern-based and rule-based approaches in a single hybrid system designed to translate Chinese cooking recipes into English. The second system (presented by Lin Quen of Peking University) likewise uses a ‘multi-engine’ architecture for general-purpose translations. In both cases it is the blending of approaches which reflects a re-orientation away from formal-theoretical paradigms towards strategies which more closely model the work of the practising translator. While Chang Baobao et al describe work on constructing and designing a Chinese-English bilingual corpus for use in translation memory systems, a paper by Li Tangqiu et al present a hybrid method for disambiguating syntactic structures and word senses in Chinese which chooses the best candidate the best candidate from multi-parse trees by comparing the degree of syntactic and semantic collocation between the words in the structure being analysed. The knowledge to do this comes from word entries stored in Hownet, a knowledge base specifically designed for NLP. In a contribution on example-based machine translation (EBMT) Kit Chunyu et al provide a comprehensive, yet concise and informative overview of the historical development of example-based MT (EMBT) and the methods used. The paper would be an excellent starting point for students needing to identify the basic issues and able to research further the algorithmic techniques that are briefly presented. Jonathan Webster et al follow this up by presenting a project at the City University of Hong Kong which is applying the example-based approach to meet the growing demand for English-Chinese translations of legal documents in Hong Kong. For data acquisition and alignment purposes semantic web technology is used to recode a 25 million word bilingual (English-Chinese) legal document corpus into marked-up, machine-readable form (in RDF/XML format).
In a section on terminology Aman Chiu and Björn Jernudd review the practical difficulties which translators and other specialists have to contend with in identifying and standardising terminological equivalents in English and Chinese with reference to the language of IT. A major problem is the existence of at least four standardisation bodies which appear not to operate with guidelines or criteria for the identification, collection and maintenance of terminological data. Pursuing this theme further, Charlotte To and Björn Jernudd discuss the terminological problems encountered by professional users of the internet (technical jargon, acronyms, pronunciation, spelling, meaning and usage) in a range of situations (business meetings, conversations, seminars, reading technical manuals, writing articles for publication on the web). They outline the strategies which users bring to bear in order to overcome obstacles of comprehension and translation.
Turning to training issues, Beverly Adab provides a detailed and informative insight into how a wide variety of software resources (notably corpora, of which the internet provides the largest single source, and search tools/concordancers such as Wordsmith and Multiconcord) are used in a postgraduate translation programme at the University of Aston. Carrie Chau Kam Hung and Irene Ip Kwok Chun report on a survey in which they assessed the perceived benefits to students of translation and interpreting of using a computer-based vocabulary-learning programme. A paper by Li Defeng provides rather negative feedback on the comprehensibility of texts generated by some well-known English-Chinese MT programs, although he also emphasises the need to develop better software and the importance of incorporating a range of IT resources in training. Paris Lau Chi-chuen supplies some interesting insights into the Chinese tradition of translation and asks whether new technology has resulted in the values of the traditional human translator being supplanted by the need for rapid decoding, speed of comprehension, partial fidelity and overall intelligibility. Comparing the translation by three Chinese-English MT systems of a very short comment by an anonymous MT researcher on the relationship between language and literature, science and machine translation, Evangeline S. P. Almberg provides an ironical comment on what computers must yet achieve before they can match the expert human translator.
In summary this volume, which unlike so many on MT is reasonably priced, provides informative perspectives on research, development, terminological and training issues. We can also expect some major advances in MT resources from China in the near future.
Derek Lewis
University of Exeter
|
