John Yen. A principled approach to reasoning about the specificity of rules. In Proceedings of the Eighth National Conference on Artificial Intelligence, pages 701--707, 1990. 15  Home/Search   Document Not in Database   Summary   Related Articles   Check

....[17] loom [15] These works allowed only rules of a restricted form: C(x) GammaD(x) where C and D are concepts. Furthermore, the rules are generally not integrated in subsumption inferences but they are just used to derive additional knowledge about concept instances. MacGregor [16] and Yen [20] describe algorithms for determining rule specificity and classification of arbitrary predicates in LOOM, which are an instance of the existential entailment problem described here. However, since subsumption in LOOM is undecidable, their algorithms are not complete either. Our work is the first ....

John Yen. A principled approach to reasoning about the specificity of rules. In Proceedings of the Eighth National Conference on Artificial Intelligence, pages 701--707, 1990. 15

....that the cost of classifying a new concept is typically logarithmic in the size of the concept taxonomy [Woods, 1991] We are hopeful that similar analysis may yield similar results for classification in our plan language. In particular, our optimism is reinforced by the normal case results of [Yen, 1990] as mentioned in Section 5.1.3. 3.2 Temporal Constraints Allen, in his influential work on maintaining knowledge about temporal intervals, enumerated a total of seven primitive relationships, plus their inverses, that might hold between any ordered pair of intervals [Allen, 1983] These ....

....antecedents of CLASP rules are composed of unary predicates (corresponding to concepts) and binary predicates (corresponding to roles) Thus they can be viewed as constraint networks. We are encouraged by Yen s analysis which found their algorithm s complexity to be polynomial in normal cases [Yen, 1990]. 5.2 Plan Recognition 5.2.1 Kautz Our plan recognition work is most closely related to that of Kautz [Kautz, 1991] Our plan recognition technique, like Kautz s, is deductive, and incorporates the use of a plan abstraction taxonomy (as well as the traditional hierarchy decomposing plans into ....

J. Yen. A principled approach to reasoning about the specificity of rules. In Proceedings of AAAI-90, pages 701--707, Boston, MA, 1990. American Association of Artificial Intelligence.

....tests is a desirable goal. Comparing metric constraints, however, is quite inexpensive so it is unlikely that this strategy would be very worthwhile. The above algorithm puts all available information to effective use. It is similar to an existing algorithm for production rule subsumption due to Yen [Yen, 1990], whose work is reviewed in Section 6.5. Our optimism is reinforced by Yen s normal case analysis for subsumption of production rules composed of description logic concepts. That algorithm s complexity was found to be polynomial in normal cases [Yen, 1990] Empirical performance analysis of our ....

....production rule subsumption due to Yen [Yen, 1990] whose work is reviewed in Section 6.5. Our optimism is reinforced by Yen s normal case analysis for subsumption of production rules composed of description logic concepts. That algorithm s complexity was found to be polynomial in normal cases [Yen, 1990] . Empirical performance analysis of our algorithm is reported in Section 4.5. Before turning to performance analysis, however, we probe more deeply into the nature of constraint network subsumption. 4.4.6 Comparison with Standard Description Logic It is worth noting some fundamental distinctions ....

J. Yen. A principled approach to reasoning about the specificity of rules. In Proceedings of AAAI-90, pages 701--707, Boston, MA, 1990. American Association of Artificial Intelligence.

....back [31] loom [28] These works allowed only rules of a restricted form: C(x) D(x) where C and D are concepts. Furthermore, the rules are generally not integrated in subsumption inferences but they are just used to derive additional knowledge about concept instances. MacGregor [29] and Yen [39] describe algorithms for determining rule specificity and classification of arbitrary predicates in LOOM, which are an instance of the existential entailment problem described here. However, since subsumption in LOOM is undecidable, their algorithms are not complete either. Our analysis of carin ....

John Yen. A principled approach to reasoning about the specificity of rules. In Proceedings of the Eighth National Conference on Artificial Intelligence, pages 701--707, 1990.

....BACK [19] LOOM [16] These works allowed only rules of a restricted form: C(x) D(x) where C and D are concepts. Furthermore, the rules are generally not integrated in subsumption inferences but they are just used to derive additional knowledge about concept instances. MacGregor [17] and Yen [22] describe algorithms for determining rule specificity and classification of arbitrary predicates in LOOM, which are an instance of the existential entailment problem described here. However, since subsumption in LOOM is undecidable, their algorithms are not complete either. ....

John Yen, `A principled approach to reasoning about the specificity of rules', in Proceedings of AAAI-90, (1990). Knowledge Representation 327 A. Levy and M.C. Rousset

....LOOM, the LOOM method dispatcher eliminates all methods whose patterns are not satisfied when their free variables are bound to the actual parameters. In LOOM, a method is preferred over another when the pattern attached to the former method specializes the pattern attached to the latter method. Yen90] describes how classifier technology can be extended to compute subsumption relations between patterns. Thus, LOOM s generalization of the method dispatching paradigm is predicated on the availability of a classifier. 7 Discussion Much of the technology described in this paper is not unique to ....

John Yen. A principled approach to reasoning about the specificity of rules. In AAAI90, Proceedings of the National Conference on Artificial Intelligence, Boston, MA, July 1990. AAAI.

....The pattern classifier organizes patterns into a lattice where more specific patterns are below more general ones, based on the definitions of terms referred to in the patterns. Using the pattern classifier, CLASP can compute a well defined specificity relation between rules during compile time [21]. Specificity is a classic conflict resolution heuristic used by many production system languages (e.g. OPS5) 22] In addition, common sense reasoning often relies on the specificity of a rule s antecedents to override conclusions drawn by more general rules when they contradict the more ....

J. Yen, "A principled approach to reasoning about the specificity of rules," In Proc. National Conf. on Artificial Intelligence, pp. 701--707, Boston, August 1990.

....A method m 2 is more specific than another method m 1 if and only if (1) they are associated with the same operator, and (2) for any invocation of the operator, if m 2 is applicable then m 1 is also applicable. Our approach for testing the specificity of methods is based on a theorem in [7] that states the sufficient and necessary condition for a conjunctive pattern (i.e. a conjuction of non negated literal) to be more specific than another one: Suppose p 2 and p 1 are two conjunctive patterns, p 2 is more specific than p 1 if and only if there exists a substitution that replaces ....

....The substitution is equivalent to a mapping that maps each variable in p 1 s condition to a variable or a constant in p 2 s condition. Based on the theorem, we have shown that an algorithm for testing the specificity of rules needs to search for a desired mapping between variables of two rules [7]. The specificity test between methods introduce one additional constraint to the mapping: the arguments of one method has to map to corresponding arguments in another method. Intuitively, it is easy to see that the existence of such a mapping is a sufficient condition that m 2 is more specific ....

[Article contains additional citation context not shown here]

J. Yen, "A principled approach to reasoning about the specificity of rules," In Proc. National Conf. on Artificial Intelligence, pp. 701--707, Boston, August 1990.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC