H. P. Grice, "What I learnt from T. C. Potts"

Structures and Categories for the Representation of Meaning Timothy C. Potts (University of Leeds) Cambridge University Press, 1994, xv + 308 pp. Hardbound, ISBN 0-521-42481-5, $54.95 This work addresses the representation problem ... how best to represent the meanings of linguistic expressions... One might call this the 'semantic form' of expressions (p. xi, italics in the original). The book begins with "three chapters in which I survey the contributions made by linguistics, logic and computer science respectively to the representation of meaning" (p. xii). These three chapters are not easy to understand, principally because of Potts's obtuse style, an example of which is that instead of saying "'either P or Q' is false if 'P' and 'Q' are both false; otherwise, it is true," he says, "we lay down that a proposition having the structure represented by 'either P or Q' is to be accounted false if a false proposition is substituted for 'P' and a false proposition for 'Q', but is otherwise to be accounted true" (p. 53). These chapters are also outdated. In particular, the chapter on computer science, discussing the work of researchers whose goals are the closest to Potts's own stated goals, is mainly a review of work as of the seventies. There are citations to several of the papers in Findler (1979), but only three to more recent research publications: Hayes (1980), Sowa (1984), and Hobbs and Shieber (1987). Perhaps the most valuable aspect of these three chapters is Potts's criticisms of some of the work he surveys. Of course, some of the problems noted have been corrected in literature that Potts hasn't yet got around to reading. By the end of the three survey chapters, Potts has introduced two techniques that he 427  Computational Linguistics Volume 21, Number 3 then develops into his own representation-- categorial grammars and graphs as representation formalisms. He takes the categorial analysis to be the prior of the two, with his graphs, which he calls categorialgraphs, being the clearer representation of sentence meaning. Unfortunately, "formalism" and "clearer" must be taken with a grain of salt. Potts never formally defines his categorial graphs, let alone gives a formal semantics for them. Although I have had extensive experience reading, interpreting, and devising graphical representations of meaning, I could not understand the details of Potts's graphs. But then, neither, apparently, can he: "The relationship between semantic and syntactic structures has not been spelled out, so that it is not fully determinate what our semantic representations represent at the syntactic level" (p. 168). The four substantive chapters are useful for the linguistic issues that they address, even if they are not useful for the representation scheme that they develop. These issues, which must eventually be faced by all knowledge representation formalisms that aspire to complete coverage of natural language include: quantifier scope; pronouns; relative clauses; count nouns, substance nouns, and proper names; generic propositions; deictic terms; plurals; identity; and adverbs. Appropriately, the book does not end on a note of claimed accomplishment, but on a note of work yet to do: "The purpose of a philosophical book is to stimulate thought, not to put it to rest with solutions to every problem ... It is still premature to formulate a graph grammar for semantic representation of everyday language... The representation problem is commonly not accorded the respect which it deserves" (p. 288). Many people agree, and have, accordingly, produced a vast literature that Potts is apparently not familiar with. (Some relevant collections are Cercone and McCalla 1987, Sowa 1991, and Lehmann 1992.) Nevertheless, Potts is still correct when he suggests that there is much work left to do.--Stuart C. Shapiro, State University of New York at Buffalo References Cercone, Nick and McCalla, Gordon (editors) (1987). The Knowledge Frontier: Essays in the Representation of Knowledge. Springer-Verlag. Findler, Nicholas V. (editor) (1979). Associative Networks: The Representation and Use of Knowledge in Computers. Academic Press. Hayes, Patrick J. (1980). "The logic of frames." In Frame Conceptions and Text Understanding, edited by Dieter Metzing, 46-61. de Gruyter, 1980. Also in Readings in Knowledge Representation, edited by Ronald J. Brachman and Hector J. Levesque, 287-295. Morgan Kaufmann. 1985. Hobbs, Jerry R., and Shieber, Stuart M. (1987). "An algorithm for generating quantifier scopings." Computational Linguistics, 13(1-2), 47-63. Lehmann, Fritz (editor) (1992). Semantic Networks in Artificial Intelligence. Pergamon Press. Sowa, John E (1984). Conceptual Structures. Addison-Wesley. Sowa, John F. (editor) (1991). Principles of Semantic Networks: Explorations in the Representation of Knowledge. Morgan Kaufmann.