P. F. Patel-Schneider, R. J. Brachman, H. J. Levesque. ARGON: Knowledge Representation meets Information Retrieval. Proceedings of the First Conference on Artificial Intelligence Applications 1984 pp 280--286 (1984)  Home/Search   Document Not in Database   Summary   Related Articles   Check

....its queries. In some approaches, the user may customize the browsing system [29, 14] or enter into a dialogue with the system in order to select better operators[5] In other approaches, the user may refine the query, as in relevance feedback systems [22, 23] or query reformulation systems [16, 13, 25, 34, 40]. Imposing structure on the library [35, 42] can also improve browsing speed and reliability. Our approach to improving browsing systems is complementary to the preceding ones. We aim to add an active component to the browsing system, so that in addition to passively supporting user directed ....

P. F. Patel-Schneider, R. J. Brachman, H. J. Levesque. ARGON: Knowledge Representation meets Information Retrieval. Proceedings of the First Conference on Artificial Intelligence Applications 1984 pp 280--286 (1984)

....could be extended so as to take into account information retrieval as we understand it. For this reason, we will not consider Darlington s work as relevant to our purposes; the same will happen, essentially for the same reasons, with the work of Frisch and Allen [15] and Patel Schneider et al. [30]) 5 to be read as the probability of # 2 . Accordingly, the central problem of this way of looking at IR becomes that of selecting the right implication connective, i.e. selecting the logic whose implication connective best mirrors relevance: the ideal logic should be the one in which P ....

Peter F. Patel-Schneider, Ronald J. Brachman, and Hector J. Levesque. ARGON: knowledge representation meets information retrieval. In Proceedings of CAIA-84, 1st IEEE Conference on Artificial Intelligence Applications, Denver, CO, 1984.

....the user s requirements. Any match on the query is therefore unlikely to be exactly what is wanted. One way to address these problems is to turn the search into an iterative process. Two popular versions of this approach are query reformulation and relevance feedback. In the 3 former, [17, 15, 27, 38, 45] a query returns an initial set of items. The user can then modify the query using information from one of these items. In response to the new query another set of items is returned and the process repeated. In relevance feedback the user critiques the items returned by the query [24, 25] The ....

P. F. Patel-Schneider, R. J. Brachman and H. J. Levesque.(1984) ARGON: Knowledge Representation meets Information Retrieval. Proceedings of the First Conference on Artificial Intelligence Applications pp 280--286

....for each of the facets. Within each facet, classification techniques are used to help users choose appropriate terms. This is very similar to the attribute value structures used in a number of frame based retrieval techniques in artificial intelligence [Brachman et al. 1991; Devanbu et al. 1991; Patel Schneider et al. 1984] except that faceted techniques use a fixed number of facets (attributes) per domain and no such restriction exists for attribute value methods [Frakes, Pole 1994] Facets are more flexible than enumerated schemes because individual facets can be re designed without impact on other facets. But ....

....key terms and phrases. Components can take on any size or form, depending on the needs of repository users. PEEL is a re engineering tool that translates Emacs Lisp files into individual, reusable, components in a frame based knowledge representation language named Kandor [Devanbu et al. 1991; Patel Schneider et al. 1984] that is used by CodeFinder to index components and create a frame based hierarchy of retrieval concepts [Henninger 1994] PEEL extracts source code definitions of functions, variables, constants, and macros from a source code file. 1 Information is extracted from each of the components and ....

[Article contains additional citation context not shown here]

Patel-Schneider, P. F., Brachman, R. J., Levesque, H. J. "ARGON: Knowledge Representation Meets Information Retrieval," Proceedings of The First Conference on Artificial Intelligence Applications (CAIA '84), pp. 280-286.

....items in both categories are retrieved. Because categories are members of exactly one category, two categories in different subtrees of the category hierarchy will return the null set. This problem is solved by allowing users to define a Boolean OR between category and attribute specifications [38], but this severely complicates the querying process by requiring users to created structured Boolean queries which have proven to be very difficult for users to generate [35] Beyond the difficulties of Boolean queries, the strict classification scheme has been observed to cause problems. People ....

P.F. Patel-Schneider, R.J. Brachman, H.J. Levesque, "ARGON: Knowledge Representation Meets Information Retrieval", Proceedings of The First Conference on Artificial Intelligence Applications (CAIA '84), 1984, pp. 280-286.

....The most important reasoning task in such a context is the determination of subsumption between concepts, i.e. whether all instances of a concept are necessarily instances of the other concept. This kind of reasoning can be employed to support such diverse applications as information retrieval [Patel Schneider et al. 1984], explainable expert systems [Neches et al. 1985] natural language processing [Webber and Bobrow, 1980; Sondheimer and Nebel, 1986] and computer configuration [Owsnicki Klewe, 1988] Based on these ideas, a number of system were built, e.g. kandor [PatelSchneider, 1984] kl two [Vilain, 1985; ....

Peter F. Patel-Schneider, Ronald J. Brachman, and Hector J. Levesque. ARGON: Knowledge representation meets information retrieval. In Proceedings of the 1st Conference on Artificial Intelligence Applications, pages 280--286, Denver, Col., 1984.

....the appropriate concepts in the concept network. It should be noted that this system did not differentiate between concepts and instances (or schema and data, respectively) but represented concepts and instances in a uniform way. Subsequently, an information retrieval system called argon [40] was developed using the concepts and ideas of the rabbit system. In contrast to rabbit, however, in argon there is a clear cut distinction between concepts and instances, and, more importantly, the system uses the technique of semantic indexing described above, which is implemented as part of the ....

P. F. Patel-Schneider, R. J. Brachman, and H. J. Levesque. ARGON: Knowledge representation meets information retrieval. In Proceedings of the 1st Conference on Artificial Intelligence Applications, pages 280--286, Denver, Col., 1984. Terminological Reasoning and Information Management 29

....says this (assuming the obvious semantics for the constructors) The answer to such a query would be a list of individuals that satisfies the conditions of the query i.e. the ones recognized by the query description. Papers such as [Tou et al. 1982] [Patel Schneider et al. 1984], Beck et al. 1989] and [Nebel and Peltason 1991] have investigated the use of DLs as query languages. DLs are particularly useful for querying knowledge bases in situations when the user is not entirely familiar with the contents or structure of the data, or when they are not entirely sure what ....

....can be organized in a subsumption lattice. These include detecting incoherent queries (ones which cannot possibly return any individuals because of the semantics of the database) and supporting the paradigm of query specification by iterative refinement, described in [Tou et al. 1982] and [Patel Schneider et al. 1984]. Data exploration involves asking very many queries, possibly by teams of people, over an extended period of time. The DL based KBMS can automatically organize this large set of queries through the subsumption relationship, thereby allowing users to find identical or similar queries asked in the ....

[Article contains additional citation context not shown here]

Patel-Schneider, P. F., Brachman, R. J., and Levesque, H. J., "ARGON: Knowledge Representation Meets Information Retrieval," Proc. First Conf. on Artificial Intelligence Applications, Denver, CO, December, 1984, pp. 280--286.

....interface could be devised which would allow for incremental construction of complex queries, exploiting the underlying structure of the object base. In an early phase of the design of bmt line, it was considered to build such an interface based on the query by reformulation approach described in [8]. The idea was abandoned at that time because the required communication between Web browsers at the user side and the bmt line Web server would have resulted in interaction too slow to make this search interface useful. The advent of Java [5] and the small applets which could run be run by ....

Peter F. Patel-Schneider, Ronald J. Brachman, and Hector J. Levesque. ARGON: Knowledge representation meets information retrieval. In Proceedings of the 1st Conference on Artificial Intelligence Applications, pages 280--286, Denver, Col., 1984.

....can find similar queries that have been asked in the past [16] Moreover, users may record observations about the query and its answer, using meta objects. ffl Query formulation by refinement: A user exploring the database can use the classification hierarchy of concepts to refine her queries [34, 39, 7]. ffl Intensional query processing: The general processing strategy of DL queries is to classify the query description with respect to the pre computed views (or saved previous queries) and then only test their instances rather than processing the entire database [7, 32] ffl Schema browsing: ....

Patel-Schneider, P. F., Brachman, R. J., and Levesque, H. J., "ARGON: Knowledge Representation Meets Information Retrieval," Proc. First Conf. on Artificial Intelligence Applications, Denver, CO, December, 1984, pp. 280--286.

....more complicated. In a presentation planning application [3] the information associated with the concepts in MSC(c) might be used to decide how to represent a given object. In a database or information retrieval application, MSC(c) can be used to index the data objects by the concepts in MSC(c) [4, 6, 50]. Query processing can then be implemented as classification of a query concept, retrieval of all objects indexed by the immediate superconcepts of the query concept in the concept taxonomy, and filtering by testing each retrieved object against the query concept. 4 Algorithmic Considerations ....

P. F. Patel-Schneider, R. J. Brachman, and H. J. Levesque. ARGON: knowledge representation meets information retrieval. In Proceedings of the 1st Conference on Artificial Intelligence Applications, pages 280--286, Denver, Col., 1984.

....department , then the description and(COURSE,at least(10,takers) all(taughtBy,all(in dept,SCIENCE DEPT) expresses this. The answer to such a query would be a list of individuals that satisfies the conditions of the query i.e. the ones recognized by the query description. Papers such as [67, 57, 8, 52] and [27] have investigated the use of DLs as query languages. DLs are particularly useful for querying knowledge bases in situations when the user is not entirely familiar with the contents or structure of the data, or when they are not entirely sure what question they should be asking. The ....

.... subsuming descriptions provides the obvious space to search for such generalizations, and therefore the system can provide a helping hand in this task, as illustrated in [3] ffl The description lattice supports the paradigm of query specification by iterative refinement, described in [67] and [57]. ffl Data exploration involves asking very many queries, possibly by teams of people, over an extended period of time. The DL based KBMS can automatically organize this large set of queries through the subsumption relationship, thereby allowing users to find identical or similar queries asked ....

[Article contains additional citation context not shown here]

P.F. Patel-Schneider, R.J. Brachman, and H.J. Levesque, "ARGON: knowledge representation meets information retrieval," Proc. First Conf. on Artificial Intelligence Applications, Denver, CO, December, 1984, pp. 280--286.

....by indexing them with key terms and phrases. Components can take on any size or form, depending on the needs of repository users. PEEL is a reengineering tool that translates Emacs Lisp files into individual, reusable, components in a frame based knowledge representation language named Kandor [8, 24] that is used by CodeFinder to index components and create a frame based hierarchy of retrieval concepts [18] PEEL extracts source code definitions of functions, variables, constants, and macros from a source code file. 1 Information is extracted from each of the components and translated into ....

....Kandor enables the expression of constraints on members of a defined frame. Frame definitions are organized hierarchically and defined through restrictions on super frames. Part of the hierarchy for the Emacs Lisp repository is shown in Figure 3. CodeFinder [18] as well as LaSSIE [8] Argon [24] and Helgon [11] have used the inferencing capabili ties of inheritance and classifi cation to ret rieve information wi th Kandor representations. Kandor s 1 At this point, the repository only contains source code functions and data structures. Nothing in CodeFinder prevents using other types of ....

[Article contains additional citation context not shown here]

P. F. Patel-Schneider, R. J. Brachman, H. J. Levesque, "ARGON: Knowledge Representation Meets Information Retrieval," Proceedings of The First Conference on Artificial Intelligence Applications (CAIA '84), 1984, pp. 280-286.

....the paradigmatic usage is a query from a user that describes an operation that she wants to see examples of, to understand in more detail, or to reuse; the LaSSIE system will retrieve instances. LaSSIE also has an interactive graphical user interface that includes a modified version of the ARGON [ 33 ] system, and the ISI Grapher [ 38 ] In this section, we present examples that illustrate how the retrieval system works. The core of the retrieval system is the classification algorithm; it simplifies the task of querying a large knowledge base. A LaSSIE query 12 is simply a description of an ....

.... individual Attd Merge Call Action 2 (CALL MERGE ACTION ATTD CAUSED ACTION) 3 (has actor Attd Monitor Process) 4 (has agent System Fabric Manager Process) 5 (has operand Generic Call) 6 (has recipient Generic Call) 7 (has environment Generic Call State) 12 The actual query language in ARGON [33], the system that LaSSIE is based on, includes negation, disjunction, etc; we illustrate just a simple query here. 8 (has result Talking State) 9 (has cause Attd Button Push) 10 (implemented by Src Ugp Attd Atd Au Im.c) 11 (calls function Signal Error Action) 12 (accesses variable ....

Patel-Schneider, P. F., Brachman, R. J., and Levesque, H. J. Argon: Knowledge Representation meets Information Retrieval. In Proc. First Conference on Artificial Intelligence Applications, 1984, pp. 280--286.

....order of other descriptions, by finding all subsuming (more general) descriptions and all subsumed (more specific) descriptions. kl one based classification systems were subsequently used in a number of interesting applications, including natural language understanding [11] information retrieval [27], expert systems [22] and more. Because of this view of frames, the research foci in the kl one family gradually diverged somewhat from those of other frame projects, which continued to emphasize typicality and defaults. Another key issue in the kl one community has been the tension between the ....

Patel-Schneider, P. F., Brachman, R. J., and Levesque, H. J., "ARGON: Knowledge Representation meets Information Retrieval," Proc. First Conf. on Artificial Intelligence Applications, Denver, CO, December, 1984, pp. 280--286.

.... cause other individuals to be reclassified, but this process is guaranteed to end because it is bounded by the number of classes and individuals in the database: every individual can move into a class at most once (since there is no removal ) Query answering follows the technique introduced in [ 25 ] : first, the query concept is itself classified with respect to the concepts in the schema; 6 then the instances of the parent concepts are tested individually to see if they satisfy the query concept. The advantage of this technique is that all instances of schema concepts that are subsumed ....

Patel-Schneider, P. F., Brachman, R. J., and Levesque, H. J. ARGON: Knowledge representation meets information retrieval. In Proc. First Conference on Artificial Intelligence Applications, pages 280--286, 1984.

No context found.

--228. PATEL-SCHNEIDER,P.F.,BRACHMAN,R.J.,AND LEVESQUE, H. J. 1984. ARGON: Knowledge representation meets information retrieval. In Proceedings of the 1st Conference on Artificial Intelligence Applications (CAIA '84). IEEE Computer Society Press, Los Alamitos, Calif.,

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