Book Reviews: Theory and Practice in Corpus Linguistics

Corpus linguistics is a hot topic, and for good reason. Text is more available than ever before. And consequently it is easier to use corpus data more effectively than it was in the 1950s, the last time that empiricism was in fashion. Corpus linguistics is such a hot area that it is already splitting up into a number of different sub-areas. Theory and Practice in Corpus Linguistics focuses on a direction practiced in much of the U.K. and Scandinavia. This work has produced a number of part-of-speech taggers and parsers based on probabilities derived from corpus data. These programs work on unrestricted texts, with reasonable accuracy and efficiency. A good example of this approach is Garside, Leech, and Sampson (1987); see Lesk (1988) for a very positive review of this book and a strong endorsement of the approach that it represents. One might contrast the view(s) held by Garside, Leech, Sampson, Lesk, and others with the A! tradition of using semantic networks and knowledge representation techniques to address lexical questions. Evens (1988), a volume from the ACL book series, is a good example of the AI approach to computational lexicography. The knowledgebased approaches tend to assume that the representation is central, and that much of the knowledge has to be entered into the system by hand, in contrast with probabilisticbased approaches where much of the knowledge is acquired through various training procedures that fit certain parameters to corpus data. Yet a third quite distinct approach can be found among lexicographers, who have recently become interested in computational issues because of the success of the COBUILD dictionary: