A Computational Theory of Writing Systems Richard Sproat (AT&T Laboratories) Cambridge University Press (Studies in natural language processing, edited by Branimir Boguraev), 2000, xviii+236 pp; hardbound, ISBN 0-521-66340-7, $59.95

Interest in the history, theory, and classification of writing systems has never been higher, and the last decade saw the publication of several worthy books on the subject, including the formidable one by Daniels and Bright (1996). At the same time, under the Unicode initiative, there has been solid progress in the definition of and, finally, the implementation of standards for computer encoding and rendering of scripts used around the world. However, the implications of this multi-lingual revolution for computational linguistics beyond the level of word-processing have not been well explored, and Richard Sproat's book A Computational Theory of Writing Systems is a welcome contribution. In particular, Sproat's observations and theories are motivated and tested by years of work at AT&T on text-to-speech systems. This is not the first or the last time that the rigor of computational application, and the massive practical testing that it allows, will come back to shape theory.