Speaking: from intention to articulation

might call the “world knowledge/deep reasoning” school of thought, which is the current paradigm in AI, engendered by Charniak’s dissertation. He introduces frames, scripts and plans, MOPS, and information formats (from the Linguistic String Project, NYU). The chapter concludes with a discussion of handling dialogue. Although there is a good discussion of some points, this chapter is somewhat disappointing because many areas of research are left out or no examples are given: classic pragmatics is not discussed at all, story and text grammars are only mentioned, MOPs are only mentioned, and no mention is made of rhetorical structure theory. Perhaps the author felt that this area was not very well settled and therefore should not be treated extensively. Graduate students, however, should be treated to open problems as well as to (purported) solutions. The language generation chapter is quite brief and treats sentence generation from a syntactic point of view. A short discussion on text generation completes the chapter. Clearly, more could be done with generation even if it is “the poor cousin” of automated language research. Grishman spends little time trying to fill in the technical linguistic background of the reader. There are short introductions to the Chomsky hierarchy of grammars and predicate logic, but the reader is mostly presumed to be conversant with linguistic theory. The book is, after all, about computational linguistics. The amount of introductory material seems reasonable relative to the expository material. For computer science graduate students, however, this account would have to be supplemented with more introductory linguistics material. The book also contains little philosophical discussion. The author does not attempt to address the problem of defining meaning, let alone the role of meaning and language for the concept of mind (a hot topic for some cognitive psychologists and philosophers). Admittedly, such discussions can become abstruse and perhaps not very useful, and Grishman typically avoids such topics. The author frequently writes short evaluations of previous efforts in an attempt to weigh the contribution of the work. The evaluations seem very fair; Grishman does not take doctrinaire stances toward different approaches to natural language, an excellent way to write an introductory text. One nice touch of the book is the pursuit of topics through several different chapters. For example, the problem of quantifier ordering is discussed both in the semantics chapter as a representation problem and in the language generation chapter. Various such themes are pursued throughout the book. In summary, this introductory text takes a stand, a point of view, toward the computational linguistics field. It can be used most profitably in classes with students who already have some linguistics background. If supplemented with additional material, it could also be used for courses populated by computer science students. The book is interestingly written with many insightful discussions, and it is the only (introductory) computational linguistics textbook that looks at the field from a linguist’s point of view.