previous item next item
 |
The British Computer Society Natural Language Translation Specialist Group |
|
Machine Translation Review
No. 15, December 2005 ISSN: 1358-8346 URL: http://www.bcs-mt.org.uk/mtreview/15/5.htm (leaf 4) E-mail: admin@bcs-mt.org.uk Books Received Reviewed by Derek Lewis |
|
The following books have been received from publishers.
Véronis, Jean (ed.) (2000). Parallel Text Processing. Alignment and Use of Translation Corpora. Dordrecht/Boston/London: Kluwer Academic Publishers. 402 pp. ISBN 0-7923-6546-1 (HB) This book evolved from the ARCADE evaluation exercise for assessing alignment systems for parallel texts (i.e. those with a translation) that began in 1995. After briefly surveying the development of parallel text processing, the volume presents chapters on alignment methodology, applications, and resources and evaluation. Carl, Michael and Andy Way (eds) (2003). Recent Advances in Example-Based Machine Translation. Dordrecht/Boston/London: Kluwer Academic Publishers. 483 pp. ISBN 1-4020-1400-7 (HB) Originating from a workshop held at the MT Summit VIII in Santiago in 2001, this volume includes sections on the foundations of example-based machine translation (EBMT), run-time approaches to EBMT, template-driven EBMT, and EBMT and derivation trees. It is the first major attempt to assemble a representative number of EBMT approaches in a single book (there are contributions from 26 different authors) and to provide an up-to-date overview of the field. Damper, R.I. (ed.) (2001) Data-Driven Techniques in Speech Synthesis. Dordrecht/Boston/London: Kluwer Academic Publishers. 316 pp. ISBN 0-412-81750-0 (HB) This book begins with a helpful introduction to text-to-speech (TTS) systems that describes early work in developing rule systems for the automatic generation of speech from phonetic symbols during the 1960s. It then moves on to focus on data-driven techniques based on neural networks that emerged in the 1980s. Four chapters are devoted to techniques for converting text into phonetic transcript, while others describe a framework for integrating various sources of linguistic knowledge, the role and importance of diphones (which cover transitions between phonemes), models of intonations, and experiments on accent and phrasing. The volume thus provides an interesting review of the variety of data-driven approaches employed in speech synthesis. Harald Baayen, R. (2001). Word Frequency Distributions. Dordrecht/Boston/London: Kluwer Academic Publishers. 333 pp. ISBN 0-7923-7017-1 (HB) This volume is intended as an introduction to the statistical analysis of word frequency distributions. These are characterised by very large numbers of rare words, which leads to strange results, including the feature that mean frequencies change as the number of observations increases. The declared aim of the book is to make techniques for the analysis of such distributions more accessible to non-specialists by presenting basic concepts and notation and describing various models with examples of applications. However, the volume is not an introduction to statistics and probability theory and non-specialist linguists will the material largely incomprehensible without a familiarity with the methods and language of these fields. A CD-ROM provides the Linux-based LEXSTATS suite of programs for carrying out statistical analyses. Junqua, Jean-Claude and Gertjan van Noord (2001). Robustness in Language and Speech Technology. Dordrecht/Boston/London: Kluwer Academic Publishers. 269 pp. ISBN 0-7923-67090-1 (HB) This book brings together selected contributions from the ELSNET-sponsored 6th European Summer School on Language and Speech Communication held in Barcelona in 1998. It considers the most promising research directions to be robust feature extraction, rapid speaker adaptation, separate modelling of phonetic and speaker-dependent information, environment-adaptive automatic speech recognition and language/task adaptation, and the combination of syntactic knowledge, global semantic analysis and pragmatic information in order to produce better language models and robust parsing algorithms. The specific topics include acoustic features and distance measure, speaker compensation in automatic speech recognition, robustness in statistical language modelling, the modelling of spontaneous speech events, regular approximation of context-free grammars, weighted grammar tools, robust parsing of word graphs, and balancing robustness and efficiency. Minker, Wolfgang, Mariani, Joseph and Alex Waibel (1999). Stochastically-Based Semantic Analysis. Dordrecht/Boston/London: Kluwer Academic Publishers. 221 pp. ISBN 0-7923-8571-3 (HB). This volume investigates the problem of automatic language understanding (NLU) for spoken language systems and proposes a parsing method that is designed to be sufficiently general and flexible so as to be easily ported to different applications, domains and human languages. After reviewing the background and research in NLU, including existing systems, a chapter introduces the applications, subject areas and natural language corpora used in the authors’ system, while others describe rule-based and stochastic NLU components. The conclusion analyses the performance of a specifically developed stochastically based frame parser using a semantic case grammar formalism that was applied to the American ATIS (Air Travel Information Services) task and ported to the French MASK (Multimodal-Multimedia Automated Service Kiosk) application and the ESST (English Spontaneous Speech Task). Stede, Manfred (1999). Lexical Semantics and Knowledge Representation in Multilingual Text Generation. Dordrecht/Boston/London: Kluwer Academic Publishers. 219 pp. ISBN. (HB) After reviewing the nature of lexical items and their representation in natural language processing, this book considers how to classify lexical variation, build domain models for natural language generation (NLG), develop levels of representation, and analyses so-called verb alternations and extensions. After describing the prototype of a system architecture for NLG, the author demonstrates how the representations in the domain model and in the lexicon can work together in the system architecture to derive different verbalisations and lexical paraphrases at sentence level. He also investigates the possibilities for extending the system into a text generator for paragraphs. The volume is an extended and revised version of a dissertation completed at Toronto and Ulm (Germany). Nirenberg, Sergei, Harold Somers and Yorick Wilks (eds) (2003). Readings in Machine Translation. Cambridge, Massachusetts: The MIT Press. i-xv and 406 pp. ISBN0-262-14074-8 (HB) This handsomely produced volume is primarily a collection of 36 hard-to-find but often-cited papers on Machine Translation. The editors describe them as ‘the “classical” MT papers that researchers and students want, or should be persuaded, to read’ and which make up the ‘communal inheritance’ of MT (for readers’ convenience a full list of the papers is included at the end of this short review). The three main sections are historical, theoretical and methodological issues, and system design. Each section is prefaced with a brief introduction. The first section contains ‘historical papers’ from the beginnings of MT (including Warren Weaver’s memorandum, Bar-Hillel’s ‘box in the pen’ article, and excerpts from the ALPAC report) up until the late 1960s. The second section looks at sublanguage, controlled input, the role of human beings in MT, the impact of linguistic approaches, the transfer/interlingua issue, and representation of meaning and knowledge. The third section is devoted to system design and includes papers on knowledge-based, statistical and example-based translation. Given the historical approach and the uncertainty of knowing what usefully to include beyond the early 1990s, this is when coverage ends. All the papers appeared originally in English and the authors admit with regret that material from Russia/USSR, Japan and France and other languages could not be included. Individual systems are not described, although references to such descriptions, and also to major textbooks, are provided. There is no doubt that reading the papers provides a fascinating and absorbing tour through several decades of MT activity. The introductions alone give crisp and clear overviews of the key developments. A particularly positive feature of this well produced, lengthy and large format book is its reasonable price (just under £36). Section 1: HISTORICAL
Section 2: THEORETICAL AND METHODOLOGICAL ISSUES
Section 3: SYSTEM DESIGN
|