DIPLOMAT: Compiling Prioritized Default Rules into Ordinary Logic Programs, for E-Commerce Applications

Rules promise to be widely useful in Internet electronic commerce. Declarative prioritized default rule knowledge representations offer the advantage of handling conflicts that arise in updating rule sets, but have as yet had little practical deployment. DIPLOMAT is a Java library that embodies a new approach to the implementation of such prioritized default rules: to compile them into ordinary logic programs (LP’s) cf. pure Prolog. We apply the approach to a newly generalized version of courteous LP’s, a semantically attractive and computationally tractable form of prioritized default rules. Compilation enables courteous LP’s functionality to be added modularly to ordinary LP rule engines, via a preprocessor, with tractable computational overhead. This takes a long step towards actual deployment of prioritized-defanlt knowledge r presentation in commercially fielded technology and applications. We give in the demo storyboard an automated example -commerce application scenario: inferencing in a 70-rule courteous LP that represents personalized pricing and promotions on a bookstore’s Web storefront. The extended version of this paper, available as an IBM Research Report~ contains the demo storyboard and technical details. Rules in E-Commerce, Challenge of Prioritized Default Rules Rules1 promise to be widely useful in Internet electronic commerce as an automatically executable specification language / programming mechanism. Rules are useful, for example, to represent: seller offerings of products and services, buyer requirements, contractual agreements, authorization policies, and more generally, many e-business policies and processes. A characteristic of this realm is that, often, rules need to be modified Copyright (~) 1999, American Association for Artificial Intelligence (www.aaai.org). All rights reserved. l (in the sense of knowledge r presentation and rule-based systems) frequently and by multiple players. Conflicts between rules often arise during updating and merging. In e-commerce applications, rules are specified by business-domain experts such as marketing managers and are modified frequently, including by merging rules specified by different people in different organizations. Another characteristic of this realm is that it is desirable for the rules knowledge representation (KR) to highly declarative: to have a semantics, independent of inferencing algorithm details, that specifies which set of conclusions are entailed by a given set of premise rules. Declarativeness aids exchange of rules between heterogeneous applications or enterprises, modification of rules, understandability of rules by humans. We are attracted by some virtues of prioritized default rules (in declarative KR): they handle conflicts, including during updating of rule sets, using partiallyordered prioritization info that is naturally available based on relative specificity, recency, and authority. In Internet e-commerce, the prioritized default expressive features are valuable especially because they greatly facilitate incremental specification, by often eliminating the need to explicitly modify previous rules when updating or merging. The overall problem we address is: how to enable prioritized default rules to be used as a widely practical knowledge representation for specification and execution of rule-based software, especially in e-commerce. Prioritized default rules are of long-standing interest in the knowledge representation (KR) community, and have received much study. However, they have as yet had little impact on practical rule-based systems and software engineering generally, and had very few deployed serious applications. One difficulty is getting the semantics right, including intuitively simple enough that non-experts in KR can feel comfortable specifying, and often repeatedly modifying, rule sets. Another difficulty is the complexity of implementing inferencing in a new KR. A third difficulty is facilitating a transition, which is best made incrementally, by builders and users of previous rulebased technology, to a new representation. We take a new overall approach to remedy these three difficulties, especially the third. From: AAAI-99 Proceedings. Copyright © 1999, AAAI (www.aaai.org). All rights reserved.