A. Borgida, "Towards the systematic development of terminological reasoners: clasp reconstructed", Proc. KR'92,  Home/Search   Document Not in Database   Summary   Related Articles   Check

....are comparatively unwieldy monolithic structures. On the other hand, CLASP supports disjunction and looping. Besides TREX, there are two projects currently extending TKR to handle plans. RAT [ Heinsohn et al. 1992 ] focuses on the representation of complex state descriptions, while PROTODL [ Borgida, 1992 ] reconstructs CLASP using natural semantics. While there have been many approaches to plan recognition, e.g. Allen and Perrault, 1980, Carberry, 1990, Cohen and Levesque, 1990, Litman and Allen, 1987, Pollack, 1990, Sidner, 1985 ] our work is most closely related to that of Kautz [ Kautz, ....

A. Borgida. Towards the systematic development of terminological reasoners: Clasp reconstructed. In Third International Conference on Principles of Knowledge Representation and Reasoning (KR'92), Cambridge, MA, 1992.

....composed nicely from binary constraints, making for a compact and facile notation. Regular expressions are comparatively unwieldy monolithic structures. On the other hand, CLASP models preconditions and effects of actions and plans, and it fully supports disjunction and looping. Recently, PROTODL [Borgida, 1992] introduced a framework for extending terminological systems with customized language constructs. This methodology was demonstrated by reconstructing CLASP in PROTODL. RAT The RAT system [Heinsohn et al. 1992] is used in the WIP project at the German Research Center for Artificial Intelligence ....

A. Borgida. Towards the systematic development of terminological reasoners: Clasp reconstructed. In Third International Conference on Principles of Knowledge Representation and Reasoning (KR'92), Cambridge, MA, 1992.

....has an underlined counterpart in normalized and completed ibm risc cpu. This structural correspondence establishes the subsumption relationship. Structural subsumption works well for the restricted languages of systems like k rep, classic, and loom, and its merits are extolled in [Borgida, 1992]. How well it can be extended for extremely expressive languages is unclear, but see [MacGregor, 1994] for an interesting proposal regarding a description classifier for the predicate calculus. A completely different approach, found notably in kris, relies on a satisfiability checking algorithm ....

....In practice, though, clasp achieves considerable leverage from the compact representation afforded by finite state machines 219 corresponding to the regular expressions. Recently, protodl introduced a framework for extending description logic systems with customized language constructs [Borgida, 1992]. This methodology was demonstrated by reconstructing clasp in protodl. 6.3.3 RAT The rat system [Heinsohn et al. 1992a] was developed at the German Research Center for Artificial Intelligence (DFKI) as part of the WIP project on automatic generation of multimedia presentations. rat was used to ....

A. Borgida. Towards the systematic development of terminological reasoners: Clasp reconstructed. In Third International Conference on Principles of Knowledge Representation and Reasoning (KR'92), Cambridge, MA, 1992.

....target 3 . An action may also specify further constraints like a location zone. Other aspects like qualitative durations may be added. 3. 2 Description Logic based representation Recent works have promoted Description Logics (DL) as formal languages for action representation [DevanbuLitman, Borgida] because they can both support the definition of taxonomic domains and provide built in classification mechanisms [Nebel] In [Weida Litman] the benefit of using DL lies mainly in the clear management of action classification rather than in DL s expression power. Works like [Borgida, DiEugenio] ....

....Borgida] because they can both support the definition of taxonomic domains and provide built in classification mechanisms [Nebel] In [Weida Litman] the benefit of using DL lies mainly in the clear management of action classification rather than in DL s expression power. Works like [Borgida, DiEugenio] show how DL can be used to represent complex compositional actions. Figure 1, together with the appendice, gives a flavour of our DL based action representation. The reader is referred to [Nebel] for a detailed introduction DL. DL helps to formalize action taxonomies. An action ....

[Article contains additional citation context not shown here]

A.Borgida, Towards a Systematic Development of Terminological Reasoners: Clasp reconstructed, KR'92, pp 259-269, 1992.

....same as. There is no way out of this dilemma in general. The benefit provided by DL research is the a priori knowledge of what combinations are undecidable, so that one is aware that complete reasoning cannot be expected in principle. We remark that the recent trend to extensible DL reasoners [4], allows specialized, possibly incomplete reasoners to be customized for applications. For example, probably the only important reasoning with keys is like (key [madeBy,serialNo,#) CAR) # (key [madeBy,serialNo) VEHICLE) for any attribute list #. The classic 2.3 release allows this kind of ....

....engine. Moreover, the correspondence to PDL allows much more complex programs to be written, describing complex local conditions under which methods can be invoked. Alternatively, one could use the plan based DL clasp [11] which can be added onto classic using the extensibility features [4]. Plans are constructed through sequence, alternation and looping from actions, with the additional benefit of checking the consistency of sequencing: preconditions of actions cannot conflict with postconditions of their predecessors. 6 Discussion We have argued that IDLs, such as that of OMG, ....

A. Borgida, "Towards the systematic development of terminological reasoners: clasp reconstructed", Proc. KR'92,

....where # is a free variable, while all(authoredBy,VENUSIAN) corresponds to #y. authoredBy(#, y) # VENUSIAN(y) The connection between these two kinds of specifications is that interpretations can be applied to both predicate calculus formulas as well as concepts (Baader, 1996; Borgida, 402 Extensible DL Representation and Reasoning TERM INTERPRETATION thing # I nothing # and( C 1 , C n ) C I 1 # . # C I n at least(n, p) d # # I p I (d) # n at most(n, p) d # # I p I (d) # n all(p, C) d # # I p I (d) # C I ....

Borgida, A. (1992b). Towards the systematic development of terminological reasoners: clasp reconstructed. In Proceedings of KR'92.

....Individual Reasoning in ProtoDL Alex Borgida Dept. of Computer Science Rutgers University http: www.cs.rutgers.edu borgida 1. Requirements This paper continues our explorations [1, 3, 2] into the implementation of extensible DL reasoners that try to maintain a model like data structure, rather than using a refutation technique 1 A DL KBMS 2 typically manages a set of named concept definitions ( classes ) including maintaining an up to date taxonomy of them organized by ....

....of magnitude larger than concept level information. The ProtoDL architecture uses three functions: Recognizes (ind,term) ConsistentW (ind,term) and InferFrom(ind,term) in order to implement the various tell operations, as well as the one that performs classification. As with concept processing [1], we use the and constructor to drive the processing, and modularize the implementation so each kind of concept constructor K has its own set of procedures: Recognizes [K] ind,term) ConsistentW [K] ind,term) and InferFrom[K] ind,term) 3 . For examples, see Figure 1. 1 Hence, the major weakness ....

A. Borgida, "Towards the Systematic Development of Terminological Reasoners: CLASP Reconstructed", in Proc. KR'92, Boston, MA, 1992, pp.259--269

....engine. Moreover, the correspondence to PDL allows much more complex programs to be written, describing complex local conditions under which methods can be invoked. Alternatively, one could use the plan based DL clasp [11] which can be added onto classic using the extensibility features [4]. Plans are constructed through sequence, alternation and looping from actions, with the additional benefit of checking the consistency of sequencing: preconditions of actions cannot conflict with postconditions of their predecessors. In many OO methods, including UML, one can provide a ....

A. Borgida, "Towards the systematic development of terminological reasoners: clasp reconstructed ", Proc. KR'92,

....Classic was used, among others, as a language for capturing the domain model, as a unifying object centered representation of information This work was supported in part by NSF grants IRI 91 19310 and IRI 9619979. The material in Section 3. 2 and the appendix have also been treated in references [8, 6]. 1 For examples, see [9, 3, 12, 14, 13] the proceedings of a series of workshops on databases and knowledge representation [20] as well as the review article [5] retrieved from archival data bases, as a query language (supporting the organization of queries themselves) and potentially as a ....

....which checks every interval (b; e) in get[dateRange] lower) to make sure that year(b) year(e) and that the month and day meet the conditions of the period. Finally, note that we have detailed elsewhere other extensions to ProtoDL, including ones for reasoning about plans [6], concept constructors missing from Classic [8] and strings [7] 3.4 Relationship to Concrete Domain extensions Although we shall address general work on extensible KR R in the conclusion, there is one specific approach involving description logics that deserve closer scrutiny at this stage, ....

A. Borgida, "Towards the Systematic Development of Terminological Reasoners: CLASP Reconstructed", in Principles of Knowledge Representation and Reasoning: Proceedings of the Third International Conference (KR'92), Boston, MA, 1992, pp.259--269

....in [2] In a nut shell, the subsumption relation can be used to detect incoherence and to organize generic ideas; and if queries are descriptions and subsumption is used as part of query processing then descriptions can be used to assert incomplete information about individuals. In previous papers [3, 1], we have argued that the expressiveness intractability trade off for DL reasoners can be most advantageously resolved by abandoning the search for an ideal DL, and by adopting a framework where one starts with a small set of description constructors, and extends them on a perapplication ....

....time, so that a significant part of the job is to reduce the amount of work needed to re check the inferences that might be applicable after an update. As an analogy, suppose that concept definitions might evolve over time; then, in addition to techniques for efficiently classifying concepts [1], a relevant research question would be to consider concept re classification) The approach presented here, and implemented in Common Lisp, is based on the structural subsumption paradigm [7] where a description D is normalized to a conjunction of independent descriptions D j , each built ....

Borgida, A. "Towards the systematic development of terminological reasoners: clasp reconstructed ", Proc. Conf. on Principles of Knowledge Representation (KR'92), Boston, MA, October 1992.

....difficult for DLs. Other kinds of semantic specification techniques have been proposed for this purpose, including non standard denotational semantics such as those in [Patel Schneider 1989a, Borgida and Patel Schneider 1992] or proof theoretic axiomatizations, such as in [Borgida 1992a, Borgida 1992b] ffl Providing an escape hatch in the language. It is possible to introduce one or more constructors in the language whose semantics are opaque for subsumption reasoning, but can still be used for recognizing individuals. For example, classic s test defined concepts are passed a Lisp or C ....

....DL reasoners which is as easy to extend as, for example, a syntax directed translation scheme used in a programming language compiler. Such extensible architectures are discussed in [Borgida and Brachman 1992, Baader and Hanscke 1991] and the methodology of providing extensions is illustrated in [Borgida 1992b] Our conclusions in this section are that The conflicting desires between expressivenss and complexity of reasoning, although very real, need not be paralyzing: there is wide variety of approaches to the problem, with the predictability of the inferences and their timing being of concern to ....

Borgida, A. "Towards the systematic development of terminological reasoners: clasp reconstructed", to appear in Proc. KR'92, Boston, MA, October 1992.

....incompleteness. As we have seen, operational definitions are relatively difficult for DLs. Other kinds of semantic specification techniques have been proposed for this purpose, including non standard denotational semantics such as those in [58, 21] or proof theoretic axiomatizations, such as in [14, 15, 59]. ffl Providing an escape hatch in the language. It is possible to introduce one or more constructors in the language whose semantics are opaque for subsumption reasoning, but can still be used for recognizing individuals. For example, classic s test defined concepts are passed a Lisp or C ....

....approach requires a modular architecture for DL reasoners which is as easy to extend as, for example, a syntax directed translation scheme used in a programming language compiler. Such extensible architectures are discussed in [17, 5] and the methodology of providing extensions is illustrated in [15]. Our conclusions in this section are that The conflicting desires between expressive languages and complexity of reasoning, although very real, need not be paralyzing: there is wide variety of approaches to the problem, with the predictability of the inferences and their timing being of ....

A. Borgida, "Towards the systematic development of terminological reasoners: clasp reconstructed ", Proc. Conf. on Principles of Knowledge Representation (KR'92), Boston, MA, October 1992.

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