Using large corpora Ed. by Susan B. Armstrong (review)

904 LANGUAGE, VOLUME 73, NUMBER 4 (1997) A dictionary of phonetics and phonology · By R· L- Trask. London & New York: Routledge, 1996. Pp. xii, 424. Terms to describe speech sounds have proliferated over the years as phoneticians and linguists have freely created and redefined terminology as they saw fit. Recommendations of the IPA are often ignored, and each phonological theory develops new or redefines old terminology to express new concepts. While this creativity enables an author to provide a logically consistent presentation within his or her own work, it also challenges readers, whether they be beginning students or seasoned linguists, to figure out what is intended. Trask's dictionary will be enormously useful in learning unfamiliar terminology and may also help writers to select existing terms to meet their needs rather than adding new ones. The dictionary encompasses not only currently fashionable phonetic and phonological terms but also those in older works, including some which T labels obsolete. Each entry has a clear definition which is usually brief but may take up to as much as two pages for such topics as 'distinctive feature' or 'phoneme'. All terms in a definition that are defined elsewhere in the dictionary are in boldface, and additionally there may be cross references to other entries. The definition may be supplemented by examples and information about the origin of the term. Especially useful are citations in most entries to publications where the concepts are discussed more fully—the list of references in the back is 20 pages long. The Southern British pronunciation of most (but not all) terms is given in IPA transcription. At times, T also provides some information about usage and may even become prescriptive, as for example when he states: 'The term "creak" is sometimes used interchangeably with creaky voice, but this usage should be avoided' (97). The recent revision of the IPA chart is included, but there are surprisingly few drawings or illustrations , given the common use of these in phonetic descriptions. Some help with phonological notation can be found in entries with the name of the notation, but there is no way to find them unless you know the name. An appendix with a list of these notations would be a useful addition to future editions. It is inevitable in an ambitious undertaking of this kind that some terms will be missing ('bark scale' and 'P-center' , for example) and that some phoneticians or phonologists will quibble about a definition. However, such instances are very rare. The print is larger than that normally found in dictionaries. While this is a boon to those with failing eyesight, it makes the volume bulkier than necessary. In light ofthe dictionary's comprehensiveness and clarity, it should be recommended to all beginning linguistics students for their professional reference bookshelf. It is, indeed, useful for any linguist since very few of us will have had the opportunity to read extensively enough to be familiar with every term included. [Frances Ingemann, University of Kansas.] Using large corpora. Ed. by Susan B. Armstrong. Cambridge, MA: MIT Press, 1994. Pp. viii, 349. The present collection reprints contributions to two special 1993 issues of Computational Linguistics which were devoted to the use of large corpora. The leitmotifuniting all papers is that with recent technological advances m hardware and the convenient availability of many and varied machine-readable corpora, it is time that the gulf—frequently widened beyond any actual need—between statisticsbased /probabilistic and rule-based/algorithmic approaches in NLP should be bridged. Kenneth W. Church and Robert L. Mercer's introduction (1-24) surveys the last 30 years in speech research, part-of-speech tagging, parsing, machine translation, and computer-aided lexicography to make this point at length. Four of the remaining eleven papers cover questions of lexicography, including collocations (Donald Hindle and Mats Rooth, Frank Smajda, James Pustejovsky et al., Ralph Weischedel et al.). Problems ansing in tagging and parsing naturalistic 'dirty data' are discussed by Ted Briscoe and John Carroll, Mitchell P. Marcus et al., and—to some extent—in the contribution by Michael R. Brent. Automatic alignment in bilingual and/or translation corpora is the focus of two contributions (William A. Gale and Kenneth W. Church...