Schmolze, J. G. and Lipkis, T. A. 1983. Classification in the KL-ONE knowledge representation system. In IJCAI-83. 330--332.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....for taxonomias such as those written in NIKL; KREME has proven effective in a number of BBN expert system efforts other than NLP and having a taxonomic knowledge base. In choosing NIKL to axiomatize the constants, one could use its built in, incomplete inference algorithm, the classifier [27]. In Janus, the classifier is used only for consistency checking when modifying or loading the taxonomic network; any concepts or roles identitled by the classifier as identical are candidates for further axiomafization. Our semantic procedures do not need even as sophisticated an algorithm as the ....

....KL TWO[31] which marries a frame system (NIKL) with propositional logic (RUP[20] Limited inference in propositional logic is the goal of KL TWO. Limited aspects of universal quantification are achieved via allowing demons in the inference process. KL TWO and its classification algorithm [27] are at the heart of the lexicalization process of the text generator Penman [28] KRYPTON [9] which marries a frame system with first order logic. The frame system is designed to be less expressive than NIKL to allow rapid checking for disjointheSS of two class concepts in order to support ....

Schmolze, J.G., Lipkis, T.A. Classification in the KL-ONE Knowledge Representation System. Proceedings of the Eighth International Joint Conference on Artificial Intelligence, 1983.

....con junction. Wh questions (who, what, etc. 3mmancls, and yes no queries are handled, end some lasses of helpful responses are produced. All three efforts employ a search procedure. In the Honeywell effort, graph matching is at the heart of the search; in the USC ISI effort, the NIKL classifier [10] is at the heart of the search; in our effort, a beam search with a cost function is used. Only our effort has been tested on applications with a potentially large search space (800 services) the other efforts have thus far been tested on applica tions with retatkeiy few services. 6. ....

Schmolz$, J.G., Lipkis, T.A. Classification in the KL-ONE Knowledge Representation System. Proceedings of the Eighth International Joint Confarenos on Artificial Intelligence, 1983.

.... existing, subclass relationships between thetin The subsumesO function [14] is a boolean function that, given two classes, determines if a subclass relationship exists between them as shown next: JTif C 2 is a C Subsumes( C , C 2) IF otherwise Its main problem is that it is not decidable [17], although this can be avoided in part restricting the expressions of the predicate definitions, expressing them in a canonical form to allow us a decidable verification. subsumesO analyses the relationship existing between the types and extensions of the input classes. This fact must be taken ....

J.G. Schmolze, T.A. Lipkis, "Classification in the KL-ONE Knowledge Representation System", Proc. of the 8th IJCAI, William Kauffmann, 1983, pp. 330-332.

....classified in the hierarchies that belong to its respective cluster. In the example considered, the derived class EMPLOYEES is included into the cluster of its base class EMPLOYEES from which it is defined. 2.3. 2 Subsumption between classes As pointed out in [Rundensteiner, 1992b] taken from [Schmolze Lipkis, 1983]: Classification is the process of taking a new (class) description and putting it where it belongs in the (class) hierarchy. In the process of automatically integrating a derived class into an object schema, or in the automatic verification of the result obtained from a manual process of ....

J. Schmolze, T. Lipkis, "Classification in the KL-ONE Knowledge Representation System," The Eigth Int'l. Joint Conf. on Artificial Inteligence, vol. 1, pp. 330332, Aug. 1983.

....attachment; default are presented in [Bra85, Pag92, RN87] among others. Examples of filters (compute with filter) procedural attachments (compute with) and default values (default) are given in the description of the class protein gene, section 5. Classification of an instance [Nap92, SL83, Mac91b] down a class hierarchy is aimed at finding the most specialized class the instance matches with. It is an inference process, because the more the class is specialized, the more there is knowledge about the individuals it groups. Hence, the lower an instance is attached to a class of the ....

J.G. Schmolze and T.A. Lipkis. Classification in the kl-one knowledge representation system. In 8th International Joint Conference on Artificial Intelligence, Karlsruhe (Germany) , 1983.

....the efficiency of the representation system. 5. Therefore, representation systems should restrict taxonomic classification to terminological definitions alone. We call thesis (5) the restricted classification thesis. These theses were promulgated in the context of the descendants of kl one [8, 37], namely nikl [41, 26, 15] the New Implementation of KL one) kl two [41] kandor [30] Doyle Patil krypton [6] and back [42, 27] While kl one itself was designed earlier, without concern for language restrictions, the designs of krypton, kandor, and back involved theoretical analyses ....

J. G. Schmolze and T. A. Lipkis. Classification in the kl-one knowledge representation system. In Proceedings of the Eighth International Joint Conference on Artificial Intelligence, pages 330--332, 1983.

...., was O(n 2 ) whereas the cost for FL was co NP hard. This result was startling because a small change in the representation language yielded a large change in complexity. FL is a subset of the language used in KL ONE, which language therefore must be at least co NP hard. The KL ONE classifier [77] ran in polynomial time because it was incomplete it overlooked certain subsumption inferences. Nebel proved that the complexity of a language FL N , comprising a subset of the BACK language, is also co NP hard [61] FL N is a different extension of Levesque and Brachman s language FL ....

J.G. Schmolze and T.A. Lipkis. Classification in the KL-ONE knowledge representation system. In Proceedings of the 1983 International Joint Conference on Artificial Intelligence, Los Altos, CA, August 1983. Morgan Kaufmann Publishers.

.... a hierarchy; a directed acyclic graph representing the non transitive links of the partial ordering, generalization hierarchy, over conceptual graphs [10,6,2,3] Levinson s earlier work used a similar data structure for organizing chemical graphs [5] The taxonomy over KL ONE concept descriptions [13] is a hierarchy. A. The Generalization Hierarchy as a Data Structure for Storing Conceptual Graphs The nodes in the generalization hierarchy are conceptual graphs and the arcs represent the non transitive ordering between the graphs. In Fig. 7 the hierarchy is given for the graphs from the ....

J. G. Schmolze and T. A. Lipkis. "Classification in the KLONE knowledge representation system," in Alan Bundy, ed., Proceedings of Eighth International Joint Conference on Artificial Intelligence, William Kaufmann Inc, 1983.

....be used to compute the Children of u. Levinson [9] noted that the intersection of the Descendants of the Parents of u, must contain the Children of u, and that object comparison outside this intersection is unnecessary. This is very important for databases of complex objects. Previous methods [8, 12] did not use this fact, and did these extra comparisons. The following method improves on previous methods of computing this intersection [5] The method in [5] noted that for an object to be in the intersection of the Descendants of Parents, then the object must have a path to each of the Parents ....

James G. Schmolze and Thomas A. Lipkis. Classification in the KL-ONE knowledge representation system. In Alan Bundy, editor, Proceedings of Eighth International Joint Conference on Artificial Intelligence, Los Altos, California, 1983. William Kaufmann Inc. 8-12th August, Karlsruhe, West Germany.

....meant by that. Definition 2 (Subsumption) Let T be a terminology. Then we say the term t 1 is subsumed by the term t 2 in T iff for any domain D and any extension function E of T it holds that E [t 1 ] E [t 2 ] Inference algorithms computing this relationship are described, for instance, in [21]. As it turns out, a complete inference algorithm, even for the small language described, is intractable [15] However, as mentioned above, we will be satisfied with a sound algorithm as long as all obvious relationships are uncovered a claim very similar to the one made by Allen about his time ....

....to determine the concepts which most accurately describe a given object, a forward inference technique called realization [12] is usually employed. Realization is very similar to classification, an inference technique used to maintain the taxonomy of concepts in the TBox according to subsumption [21]. In fact, realization can be viewed as abstraction generating a description in terms of the TBox followed by classification of this description (cf. 23] A first approximation to the implementation of this inference could be realized as follows. After a new assertion about an indivdual ....

James G. Schmolze, Thomas Lipkis, "Classification in the KL-ONE Knowledge Representation System, " Proc. 8th IJCAI, Karlsruhe, 1983, pp. 330--332.

....] The major strength of term subsumption systems is the reasoning capabilities offered by a classifier. The classifier is a special purpose reasoner that automatically infers and maintains a consistent and accurate taxonomic lattice of logical subsumption relations between concepts [7]. Based on such inferential power, term subsumption systems tidily handle the pattern matching problem of recognizing John as a successful father, given facts such as John is a male person , John has two children , Philip is John s son , Angela is John s daughter , both Philip and Angela are ....

....no unary condition in the pattern implies (College Graduate w) even though it is implied by the pattern based on the definition of Successful father. Examples of normalized rules can be found in Figure 9. Normalizing a pattern is analogous to completing a concept definition in KL ONE s classifier[7]. Both of them attempt to compute the deductive closure of the objects to be classified before actually classifying them for the same reason: to gain efficiency for the subsumption test. The actual algorithm for normalizing patterns depends on the language used for defining terms. The rationale ....

J. Schmolze and T. Lipkis, "Classification in the KL-ONE knowledge representation system," In Proceedings of the Eighth International Joint Conference on Artificial Intelligence, pp. 330--332. IJCAI, 1983.

....namely the subsumption relation. In an early attempt to formalise the syntax and semantics of nikl Schmolze and Israel [1983] also describe part of its classifier and in particular the subsumption algorithm. To my knowledge this is the earliest formal account of a kl one based formal13 ism. Schmolze and Lipkis [1983] present a more informal account of nikl. It should be noted that the papers Schmolze and Israel [1983] and Schmolze and Lipkis [1983] actually deal with nikl, the new version of kl one see Schmolze [1989b, p. 12] A term frequently used in the literature but hardly ever explained is (role) ....

....describe part of its classifier and in particular the subsumption algorithm. To my knowledge this is the earliest formal account of a kl one based formal13 ism. Schmolze and Lipkis [1983] present a more informal account of nikl. It should be noted that the papers Schmolze and Israel [1983] and Schmolze and Lipkis [1983] actually deal with nikl, the new version of kl one see Schmolze [1989b, p. 12] A term frequently used in the literature but hardly ever explained is (role) fillers . A definition is given in Schmolze and Israel [1983, p. 34] the filler of a role is interpreted as an element in the range of ....

Schmolze, J.G., and Lipkis, T. [1983]. Classification in the klone Knowledge Representation System. International Joint Conference on Artificial Intelligence, 330-- 332.

....i.e. determining the most specific set of generic concepts which describe it, has sometimes been referred to distinctly as realization [Mark, 1981] The classification process can be automated with reasonable efficiency. Schmolze and Lipkis formally specify the classification algorithm in KL ONE [Schmolze and Lipkis, 1983]. NIKL s classifier is described in [Robins, 1986] and LOOM s is presented in [MacGregor, 1988] Automatic classification is useful for incremental construction of a taxonomy, enforcing semantics, type checking, and pattern matching. 3.1.3 Computational Complexity A seemingly insignificant ....

J. G. Schmolze and T. A. Lipkis. Classification in the kl-one knowledge representation system. In Proceedings of the International Joint Conference on Artificial Intelligence, pages 330--332, Karlsruhe, West Germany, 1983.

....be members of A. For example, in our knowledge base, Male(X) implies Person(X) for all individuals X; hence, Person subsumes Male. An important feature of classification based systems such as LOOM is their ability to compute subsumption relationships between concepts and relations. A classifier [Schmolze and Lipkis, 1983] Expression Interpretation e [ e] primitive (concept) a unique primitive concept :primitive (relation) a unique primitive relation ( and C1 C2 ) x: C1 ] x) C2 ] x) and R1 R2) xy: R1 ] x; y) R2 ] x; y) at least 1 R) x: 9y: R] x;y) exactly 1 R) x: 9y: R] x;y) 8yz: ....

James Schmolze and Thomas Lipkis. Classification in the KL-ONE knowledge representation system. In Proceedings of the Eighth International Joint Conference on Artificial Intelligence, IJCAI, 1983.

....Agency under contract no. DABT63 91 C 0025. The views and conclusions contained in this document are those of the authors and should not be interpreted as representing the official policies, either expressed or implied, of ARPA or the U.S. Government. 1 Introduction A description classifier [Schmolze Lipkis, 1983] is a central component of many modern knowledge representation (KR) languages. Classifiers provide many useful services within a unified framework, including categorization of new concepts, recognition of instances, and answering of queries about the state of a knowledge base. An undesirable ....

Schmolze, J. G., & Lipkis, T. A. (1983). Classification in the KL-ONE knowledge representation system. In Proceedings of IJCAI-83.

....to further exploit classification technology based on structured inheritance networks. Applications using this technology take a description of an item and use it to insert the item into the appropriate part of a network. Significant work has already been done in both automatic classification [Schmolze83] and in interactive classification [Finin86] By applying this work to the RLF, we may be able to further automate portions of the component installation and retrieval tasks. Similarly, machine learning research may help automate network construction. By starting with basic descriptions and ....

J. Schmolze and T. Lipkis, "Classification in the KL-ONE Knowledge Representation System," Proceedings Int. Joint Conf. on Artificial Intelligence, Karlsruhe, West Germany, 1983, pp. 330-332.

No context found.

Schmolze, J. G. and Lipkis, T. A. 1983. Classification in the KL-ONE knowledge representation system. In IJCAI-83. 330--332.

No context found.

Schmolze, J. G., and Lipkis, T. A., "Classification in the KL-ONE Knowledge Representation System, " in The Int. Joint Conf. on Artificial Intelligence, Aug. 1983, vol.1, pp. 330 -- 332.

No context found.

SCHMOLZE, J.G. & LIPKIS, T.A. (1983). Classification in the KL-ONE knowledge representation system. In A. BUNDY, Ed. Proceedings of Eighth International Joint Conference on Artificial Intelligence, Los Altos, California. William Kaufmann Inc. 8-12th August, Karlsruhe, West Germany.

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