may be used either t o c o m p u t e outputs from given inputs, or to c o m p u t e which inputs w o u l d generate a given o u t put. Though the paper o n l y sketches code, it references SeeLog as an i m p l e m e n t a t i o n that partly satisfies its specification for a "structural m a p p i n g " (p. 176) b e t w e e n database objects and their graphical representations. The v o l u m e has o t h e r papers w o r t h mentioning, such as the E r i k s s o n - J o h a n s s o n - T a r n l u n d paper on a derivation editor, the McCord c o n t r i b u t i o n on natural language processing, and the Fuchi o v e r v i e w of fifth generation s y s tems. On the whole, the papers in this v o l u m e are clearly written; address m o s t l y standard issues in logic p r o g r a m ming; and are geared t o w a r d the k n o w l e d g e a b l e public rather than the researcher. It is notch b e l o w the C l a r k Tarnlund volume, but still w o r t h having. intentions} The prefaces of individual articles recall these c o n cerns as t h e y summarize w h a t follows, in addition to referencing o t h e r chapters that pertain to the w o r k at hand. Only three chapters apart from the introduction are works of either editor, very slight s e l f - p r o m o t i o n given their renown. Print quality overall is vastly superior to the s p o t t y copies c u s t o m a r i l y obtained f r o m univeristy graphics arts services. One of the older papers exhibits slightly blurred type. Surprising flashes of candor and w i s d o m can be taken out of c o n t e x t if one desires, as w h e n Minsky brings t o g e t h e r several issues 'by pretending t o have a unified, coherent theory', and Moore observes that 'better control of the d e d u c t i o n process was needed'. The collection, while far f r o m exhaustive, provides great o p p o r t u n i t y for philosophical m u s i n g by the reader. In one sense, KR began w h e n Quillian a t t e m p t e d to model language concepts with a n e t w o r k of nodes i n t e r connected via associative links to represent d i c t i o n a r y definitions. As t h e field evolved, its practitioners gradually realized t h a t the distinction of syntax versus sematics familiar f r o m d e v e l o p m e n t of p r o g r a m m i n g languages a p plies equally to KR structures. Writings of Woods and others e l o q u e n t l y explain that the semantics of a n e t w o r k language m u s t specify precise meanings for pointers & nodes, predicates & propositions, w o r d - s e n s e s & c a s e - relationships, l a n g u a g e - d e p e n d a n t primitives, conceptual units and t h e i r interrelations. Technical e v o l u t i o n is also e v i d e n t in t h e progression from McCarthy's "Advice Taker" specification to Etherington and Reiter's explanation of w h y "spreading activation" based on shortest paths does NOT always w o r k properly { w h i c h fact others have discovered the hard way}. A n o t h e r thread of semiotic d e v e l o p m e n t concerns the role of formal logic in KR. A m o n g interesting artifacts the reader e n c o u n t e r s are included a n o n - c o m p o s i t i o n a l semantic network, and a KR system in first o r d e r logic which lacks a t h e o r e m prover. The struggle to inject semantics into KR m a y never end. Even in the midst of a lucid e x a m i n a t i o n of c o n c e p t structure in semantic nets { p 2 0 9 } one can observe p r o p a g a t i o n of a misleading d e n o t a t i o n in a w e l l - k n o w n example, w h e r e use of the label "arch" is c o n c e p t u a l l y plausible but linguistically void. Webster's definition of "arch" m e n t i o n s "a curved structure, as of masonry, t h a t supports the w e i g h t of material o v e r an open space, or any similar structure, as a m o n u m e n t " . The "lintel and jamb" structure actually represented is rich in potential for "near misses", but c o m p l e t e l y lacks any implication of c u r v a t u r e or structural support of objects o t h e r than its o w n components. The t r i u m p h of this curveless "arch" illustrates h o w easily k n o w l e d g e representation can be w h a t you make of it. Readings in Knowledge Representation Ronald J. B r a c h m a n and Hector J. Levesque, editors. ISBN 0 - 9 3 4 6 1 3 - 0 1 - X M o r g a n K a u f m a n n Publishers 95 First S t r e e t Los Altos, CA 94022 Reviewed by J. Elliott Smith This book is a c o l l e c t i o n of basic papers in the topic area grouped in six broad categories t o g e t h e r with a b i b l i ography. Each article is prefaced by a brief discussion of its historical c o n t e x t and relation to o t h e r works appearing in the volume. The editors" carefully considered selections could serve as core readings of a graduate course in KR, or a s sist c o m p r e h e n s i v e s e l f - s t u d y of the field. Many of the papers d e m o n s t r a t e that the original exposition of an idea is clearer than any e l a b o r a t i o n which follows. Comparison of the 31 articles published o v e r an i n t e r val of 18 years suggests that selection of an a p p r o p r i a t e model is m o r e i m p o r t a n t than perfecting the m e c h a n i s m which implements it. At least one a u t h o r presents pure speculation, but the m a j o r i t y discuss concrete i m p l e m e n tations { w i t h their a t t e n d a n t warts}, making the t e x t v a l u able as a guide for researchers wishing to avoid the errors of the past, or r e - e x p l o r e a b a n d o n e d directions of inquiry. Several of the m o n o g r a p h s are t r o u b l e s o m e to obtain {and perhaps more often cited than r e v i e w e d } ; their j u x taposition provides a valuable perspective. The editor's introduction explains h o w recurring issues in KR relate to its three essential ingredients: a representation language, automatic inference, and specific information. Issues considered in the collection include: * adequacy of representation {expressiveness; r e a s o n ing efficiency}; * e p i s t o m o l o g y {selection of a set of primitive c o n cepts; reflective or "meta" k n o w l e d g e ; and handling of incomplete, incorrect or counterfactual knowledge} * n o n - s t a n d a r d logic { i m p e r a t i v e versus declarative assertion; default knowledge; n o n - d e d u c t i v e and n o n - m o n o t o n i c reasoning} * arities b e y o n d verity {procedural, analogical and probabalistic representations} * questions of substance {fluid objects; causality and time; propositional attitudes, ie, beliefs and SIGART N e w s l e t t e r , O c t o b e r 1986, N u m b e r 98 Page T4

/lp/association-for-computing-machinery/book-review-readings-in-knowledge-representation-edited-by-ronald-j-24IPDPf6PE