Briefly Noted

Briefly Noted

Referential Communication Tasks George Yule

Mahwah, NJ: Lawrence Erlbaum Associates (Series on Second Language Research: Theoretical and Methodological Issues, edited-by Susan Gass and Jacquelyn Schacter), 1997, ix+125 pp; hardbound, ISBN 0-8058-2003-5, $36.00; paperbound, ISBN 0-8058-2004-3, $17.50

Referential Communication Tasks offers, in a small format (125 pages), a general survey of experimental settings and scientific back­ground on communication studies centered on specific activities in which the speakers are observed for their ability to achieve tasks rather than solely for their linguistic produc­tions. It thus offers a basic and introductory insight to research and experiments that put the notion of reference at the core of their interest. Here the focus is generally on ex­periments in the broad field of L2 studies. Though the author does not really insist on the theoretical backgrounds of the different studies being presented, it is made very clear that the language that is being described is that spoken spontaneously in situations (es­pecially in L2 contexts) in which speakers might not even be able to correctly designate the objects they want to refer to and therefore need to develop linguistic and other commu­nicative strategies to circumvent their diffi­culties. It is thus important to remember that reference is not simply "some kind of connec­tion between words and things" but rather, as stated in the first chapter of the book ("Overview"), "an action through which a speaker (or writer) uses linguistic forms to enable a listener (or reader) to identify some­thing."

Apart from the second chapter ("The de­velopment of (LI) referential communica­tion"), which is too short to be really infor­mative, the book is centered on three main sections, dedicated respectively to the main trends, the experimental settings, and the study viewpoints associated with referen­tial communication tasks. First, Chapter 3 ("Principles and distinctions") lists the main parameters—from the point of view of what is to be elicited as well as the task format— that are to lead an experimenter towards the design of a specific experimental set­ting adapted to his research interest. The au-

thor then presents (Chapter 4, "Materials and procedures") in a rather exhaustive way the different kinds of situations that have been used in the literature to conduct referential task studies, ranging from object identifica­tion to instructional map tasks (which are well known in the computational linguistics community). Finally, Chapter 5 ("Analytic frameworks") quickly surveys the possible viewpoints that the researcher may take to consider the linguistic productions resulting from the different experimental frameworks presented in the book, insisting on commu­nication strategies, interactive repair mecha­nisms, and communicative outcomes. Even though the book is far from present­ing a theoretical insight into the field, as a whole it provides a wide and extensive overview of experimental materials and prin­ciples that, considered from the point of view of computational linguistics, would be of in­terest to anyone interested in dialogue stud­ies in which reference plays a central role. Since the book seems to be intended to pro­vide the basics of some research methodol­ogy, one may regret that the literature that is mentioned contains so few associated ba­sic readings, which could have been most valuable for young researchers starting in the field of communication research. Still, its small format makes it worth reading and even to have within reach for those of us who at some point or another have to deal with, or even produce, experimental data. —Laurent Romary, Loria-CNRS

(University of Cambridge and the University of Sheffield)

Cambridge, MA: The MIT Press, 1996, vii+352 pp; hardbound, ISBN 0-262-11225-6, $35.00

We use vague language without difficulty. Yet, if asked whether a vague word ap­plies to a particular object, we might be at a loss. Is that man tall? If he is 195 cm, clearly yes. If he is 165 cm, clearly no. But in many intermediate cases, we simply can­not say. Vagueness also presents theoretical problems, among them the so-called sorites paradoxes. It would appear that a man who is a fraction of a millimeter shorter than a tall man is tall. But then, with repeated ap­plication of this principle, it seems to follow, paradoxically, that since a 195 cm man is tall, so is a 165 cm man.

The difficulties surrounding vagueness have ramifications within artificial intelli­gence. We might wish to include vague rep­resentations in knowledge bases, and in com­putational linguistics we will certainly have to come to terms with vagueness. Ques­tions immediately arise about how to rep­resent vague sentences and what to infer from them. To answer these questions, an ac­count of the meanings of vague sentences is needed.

Rosanna Keefe and Peter Smith have com­piled a collection of classic papers on vague­ness. Three ancient writings provide histor­ical perspective; fourteen others chart the development of the contemporary discus­sion of vagueness throughout this century. Contributions that might be of special in­terest to researchers in artificial intelligence include Kit Fine's "Vagueness, truth and logic," which sets out the supervaluational approach to vagueness. This approach is sup­posed to account for vagueness while pre­serving classical logic. Kenton Machina de­fends the approach to vagueness through fuzzy logic that was pioneered by Zadeh and Goguen. Dorothy Edgington criticizes this approach and suggests an alternative many-valued treatment. Max Black explores the idea that the applicability of a predicate to an object is a matter of degree, determined by the applicability judgments of the language community at large. Carl Hempel provides a pointed criticism of his proposal. Other selec­tions are philosophically rich but might only indirectly address the concerns of the artifi­cial intelligence community.

Keefe and Smith contribute a substantial introductory chapter that expounds and eval­uates these and other approaches. They dis­cuss the sorites paradox and its significance for logical accounts of vagueness; they de­scribe the epistemic view, according to which vague predicates are sharply defined but speakers are necessarily uninformed about their boundaries; they evaluate the various alternatives to this view, including super­valuational accounts, three-valued accounts, and multivalued accounts; and they outline the recent debate over the existence of vague objects. The chapter extensively refers to relevant material that does not appear in the anthology.

This balanced and comprehensive collec­tion will be a standard reference for many years to come.—Alice Kyburg, University of Wisconsin, Oshkosh

(Signiform) "The ThoughtTreasure platform was created with the long-term goal of human-level com­prehension in mind. It brings together much of the research and experience in linguis­tics and artificial intelligence to date, serving as a foundation for future work. The plat­form contains 135,000 lines of ANSI C code and concise-format database entries repre­senting over 20,000 concepts associated with over 50,000 English and French words and phrases. It includes a syntactic and seman­tic parser, an English and French genera­tor, and a chatterbot front end enabling the user to converse with the program. Thought-Treasure, including the source code, is avail­able for free on the Internet. It runs on both PCs and Unix systems."—From theauthor's in­troduction

New York: Signiform(www.signiform.com),1997, x+343 pp; spiralbound, ISBN 0-9660746-0-2, $50.00

Richard K. Larson, David S. Warren, Juliana Freire de Lima e Silva, O. Patricia Gomez, and Konstantinos Sagonas

(State University of New York, Stony Brook and Catholic University of Leuven) "Semantica is ... a new software application that allows the user to explore the seman­tic structure of language in an engaging, in­teractive way. The program, which was pro­duced as part of a National Science Foun­dation initiative for improving linguistics instruction, is designed to be used with [the authors'] Syntactica [The MIT Press, 1996], a tool for studying natural language syntax. Se­mantica provides a simple graphical interface for creating semantic theories, viewing the truth conditions that those theories assign to phrase markers created in Syntactica, and for testing those truth conditions in a pictorially represented world."—From the publisher's an­nouncement

Cambridge, MA: The MIT Press, 1997, xiii+179 pp and 3.5-inch diskettes for PCs and NeXTStations running NeXTStep 3.2 or higher; hardbound, ISBN 0-262-62117-7, $25.00

Readings in the Philosophy of Language Peter Ludlow (editor)

(State University of New York, Stony Brook) "Throughout the history of ideas, various branches of philosophy have spun off into the natural sciences, including physics, bi­ology, and perhaps most recently, cognitive psychology. A central theme of this collec­tion is that the philosophy of language, or at least a core portion of it, has matured to the point where it is now being spun off into linguistic theory. Each section of the book contains historical (twentieth-century) read­ings and, where available, recent attempts to apply the resources of contemporary lin­guistic theory to the problems under discus­sion. This approach helps to root the nat­uralization project in the leading questions of analytic philosophy. Although the older readings predate the current naturalization project, they help to lay its conceptual foun­dations. The main sections of the book, each of which is preceded by an introduction, are 'Language and Meaning,' 'Logical Form and Grammatical Form,' 'Definite and In­definite Descriptions,' 'Names,' 'Demonstra­tives,' and 'Attitude Reports'."—From thepub-lisher's announcement

Cambridge, MA: The MIT Press, 1997, xviii+1073 pp; paperbound, ISBN 0-262-62114-2, $35.00

(University of California, Los Angeles) "Ways of Scope Taking is concerned with syn­tactic, semantic, and computational aspects of scope. Its starting point is the well-known but often neglected fact that different types of quantifiers interact differently with each other and other operators. The theoretical ex­amination of significant bodies of data, both old and novel, lead to two central claims. (1) Scope is a by-product of a set of dis­tinct logical form processes; each quantifier participates in those that suit its particular features. (2) Scope interaction is further con­strained by the semantics of the interacting operators. The arguments are developed us­ing minimalist syntax, generalized quantify [sic] theory, discourse representation theory, and algebraic semantics.

Dordrecht: Kluwer Academic Publishers (Studies in Linguistics and Philosophy, edited by Gennaro Chierchia, Pauline Jacobson, and Francis J. Pelletier, volume 65), 1997, xxi+466 pp; hardbound, ISBN 0-7923-4446-4, $170.00, £99.00, Dfl 270.00;paperbound, ISBN 0-7923-4451-0, $49.00, £29.00, Dfl 90.00

"The contributors (Beghelli, Ben-Shalom, Doetjes, Farkas, Gutiérrez Rexach, Hon-coop, Stabler, Stowell, Szabolcsi, and Zwarts) make tightly related theoretical assumptions and focus on related empirical phenom­ena, which include the direct and inverse scope of quantifiers, distributivity, negation, modal and intensional contexts, weak is­lands, event-related readings, interrogatives, wh/ quantifier interactions, and Hungarian syntax. An introduction to the formal seman­tics background is provided."—From the pub­lisher's announcement

(Technical University of Vienna) "The book contains a collection of eight sur­vey papers written by some of the most ex­cellent researchers in foundations of knowl­edge representation and reasoning. It covers topics like theories of uncertainty, nonmono­tonic and causal reasoning, logic program­ming, abduction, inductive logic program­ming, description logics, complexity in arti­ficial intelligence, and model-based diagnosis. It thus provides an up-to-date coverage of recent approaches to some of the most challenging problems underlying knowledge representation and artificial intelligence in general."—From the publisher's announcement

Stanford, CA: CSLI Publications and The European Association for Logic, Language and Information, 1996, xvi+318 pp; distributed by Cambridge University Press; hardbound, ISBN 1-57586-057-0, $59.95; paperbound, ISBN 1-57586-056-2, $22.95

The contents of the volume are as follows: "Non-standard theories of uncertainty in plausible reasoning" by Didier Dubois and

Henri Prade "Probabilistic foundations of reasoning with conditionals" by Judea Pearl and Moisés

Goldszmidt "Foundations of logic programming" by

Vladimir Lifschitz "Abductive theories in artificial intelligence"

by Kurt Konolige "Inductive logic programming" by Stefan

Wrobel

"Reasoning in description logics" by Francesco M. Donini, Maurizio Lenzerini, Daniele Nardi, and Andrea Schaerf

"Artificial intelligence: A computational per­spective" by Bernhard Nebel

"The consistency-based approach to auto­mated diagnosis of devices" by Oskar Dressier and Peter Struss