R. J. Brachman. What IS-A is and isn't: An analysis of taxonomic links in semantic networks. IEEE Computer, 16(10):30--36, 1983.  Home/Search   Document Not in Database   Summary   Related Articles   Check

.... change [11] From the standpoint of semantics, Brachman describes several meanings of the isa relation that may exist between two generic concepts in semantic networks (subset superset, generalization specialization, kind of, conceptual containment, role value restriction, set prototype) [4]. He also suggests using those semantic subcomponents as the primitives of a representation system. In practice, taxonomic knowledge is complex and remains partially intuitive in many existing ontologies. This may lead to ruptures in knowledge representation, and thus impair the capability of ....

Brachman, R.J. What Is-a Is and Isn't - an Analysis of Taxonomic Links in Semantic Networks. Computer, 16 (10). 30-36.

....making models confusing and difficult to reuse or integrate. Clearly, insights into how to properly construct a taxonomy are useful. Many previous efforts at providing these insights have focused on the semantics of the taxonomic relationship (also called is a, class inclusion, subsumption, etc. [2], on different kinds of relations (generalization, specialization, subset hierarchy) according to the constraints involved in multiple taxonomic relationships (covering, partition, etc. 28] on the taxonomic relationship in the more general framework of data abstractions [7] or on structural ....

R. Brachman, What IS-A is and isn't: an analysis of taxonomic links in semantic networks, IEEE Computer 16 (10) (1983) 30-36.

....IRI 8819624 and in part by the Advanced Research Projects Agency of the Department of Defense under Office of Naval Research contract N00014 86 K 0180. 1989 Massachusetts Institute of Technology. I Introduction relationships between classes [Bobrow and Winograd, 1977] Fahlman, 1979] [Brachman, 1983], Brachman et al. 1983] Taxonomic hierarchies without defaults or exceptions are semantically equivalent to a collection of formulas in first order predicate calculus. Designers of knowledge representation lan guages have argued that there are computational advantages to representing facts as ....

....and in part by the Advanced Research Projects Agency of the Department of Defense under Office of Naval Research contract N00014 86 K 0180. 1989 Massachusetts Institute of Technology. I Introduction relationships between classes [Bobrow and Winograd, 1977] Fahlman, 1979] Brachman, 1983] [Brachman et al. 1983]. Taxonomic hierarchies without defaults or exceptions are semantically equivalent to a collection of formulas in first order predicate calculus. Designers of knowledge representation lan guages have argued that there are computational advantages to representing facts as taxonomic relationships ....

R. J. Brachman. What is-a is and isn't: An analysis of taxonomic links in semantic networks. IEEE Computer, 16(10):30-36, October 1983.

....we could also have that elephant mammal, leaving it under specified whether mammal animal or vice versa. each other, i.e. that there is a relation between concepts (which we denote ) and that this relation is a partial order. We don t say that such requirements shouldn t be stated (see e.g. [1, 7]) however, it is outside the scope of this paper where we focus on the mathematical properties of taxonomies. Actually, we are now able to specify taxonomies, simply by use of the subsumption relation. However, we prefer to introduce two operations on concepts, making the language far more ....

Ronald J. Brachman. What IS-A Is and Isn't: An Analysis of Taxonomic Links in Semantic Networks. IEEE Computer, 16(10):30--36, 1983.

....making models confusing and difcult to reuse or integrate. Clearly, insights into how to properly construct a taxonomy are useful. Many previous efforts at providing these insights have focused on the semantics of the taxonomic relationship (also called is a, class inclusion, subsumption, etc. [3], on different kinds of relations (generalization, specialization, subset hierarchy) according to the constraints involved in multiple taxonomic relationships (covering, partition, etc. 33] on the taxonomic relationship in the more general framework of data abstractions [9] or on structural ....

Brachman, R. 1983. What IS-A Is and Isn't: An Analysis of Taxonomic Links in Semantic Networks. IEEE Computer, 16(10): 30-36.

....within a Datalog framework. This paper deals with aspects that are complementary to type inferencing and method inheritance in that we concentrate on exploiting aggregation and classification knowledge. Generalization is seen as a logical implication between class membership predicates (see [BRAC83] and section 7) We therefore use an earlier and simpler object model that was first presented in [STAN86] and forms a subset of the knowledge representation language Telos [MBJK90] In this section, the formalization is briefly summarized as far as it is needed later on. 6 Definition 3 Let ....

Brachman,R.J. (1983). What IS-A is and isn't: an analysis of taxonomic links in semantic networks. IEEE Computer 16(10), Oct. 1983.

....A , respectively. In the last example, the duplicate key will open the door only if it is well made, and so the conclusion seems less certain than in the rst example. According to Brachman, understanding what is on either end of the link is also the key to understanding the import of the link [9]. This is where Aristotelian logic has something to o er to conceptual integrity over and above rigorous formalism. 1.1 Overview of Aristotelian Logic Aristotle s logical works are commonly referred to as the Organon (Greek, for tool and reason is the tool of tools) The Organon is divided by ....

....This type of is a relation that carries structure between structured descriptions is one of the most radical departures from representation schemes based on standard predicate logic. Almost all of the other is a relations are easily expressed in standard quanti cational languages. [9] This type of is a relation is at the core of Aristotelian logic, and here we wish to investigate how the object oriented programming notion of class is similar to the notion of species . Now, to say the notion of species , without quali cation of context, could easily lead us to an ....

[Article contains additional citation context not shown here]

Brachman, R. J. What is-a Is and Isn't: An Analysis of Taxonomic Links in Semantic Networks. IEEE Computer 16, 10 (October 1983), pp. 30 - 36.

....structured taxonomies have the opposite effect, making models confusing and difficult to reuse or integrate. Many previous efforts at providing some clarity in organizing taxonomies have focused on the semantics of the taxonomic relationship (also called is a, class inclusion, subsumption, etc. [3], on different kinds of relations (generalization, specialization, subset hierarchy) according to the constraints involved in multiple taxonomic relationships (covering, partition, etc. 23] on the taxonomic relationship in the more general framework of data abstractions [7] or on structural ....

Brachman, R. 1983. What IS-A Is and Isn't: An Analysis of Taxonomic Links in Semantic Networks. IEEE Computer, 16(10): 30-36.

....eds. http: sunsite.informatik.rwth aachen.de Publications CEUR WS Vol 18 I will focus here on a single design practice, which is the main responsible of most semantic difficulties: IS A overloading. Notice that I do not refer with this expression to the old debate about the semantics of IS A (Brachman, 1983), since I assume a standard set inclusion semantics for it. Rather, I will discuss the cases where this standard semantics turns out to be violated, if we carefully analyze the ontological nature of the arguments. I will consider a number of bad practice cases, proposing a way to simplify the ....

Brachman, R. 1983. What IS-A Is and Isn't: An Analysis of Taxonomic Links in Semantic Networks. IEEE Computer, 16(10): 30-36.

....more attributes on the elements of the subclass. It is more difficult to define inheritance concepts on bit strings whose interpretation is environment dependent. In mathematics, the integers are a subset of the reals; an integer is a real #################################### 26 Brachman [Bra83] correctly notes that inheritance as defined by the IS A construct is really little more than a convenient syntactic shorthand for incrementally creating subclasses. We, too, will treat it in just this fashion. This way of incrementally creating subclasses has also been called specialization ....

R. J. Brachman, What IS-A Is and Isn't: An Analysis of Taxonomic Links in Semantic Networks, COMPUTER 16,10 (Oct. 1983), 30-36.

....programming, and non monotonic reasoning systems. An unsatisfactory way of achieving this is to allow instances to be maximally specific (or leaf) sorts. The problems of mixing class and instance (i.e. subset vs. element) links in hierarchies were clearly identified by Woods [18] and Brachman [4]. Another unsatisfactory solution is to create new sorts which denote single elements, because sorts are declarative in nature whereas individuals are assertional. We propose an elegant generalization of equality constraints as a formal means of specifying and maintaining instance level ....

....set of individuals (the subset of the universe which contains the individual) If the exact individual denoted is unknown, the set represented by a variable is neither empty nor a singleton. The distinctions between sorts and individuals (or declarational vs. assertional relations) is described in [4], and the need to distinguish between subsort (i.e. isa subsort of) relations and member (i.e. isa instance of) relations is justified. Thus, we cannot intermix the sort hierarchy and individuals (where individuals might be seen as minimal sorts or leaves of the hierarchy) In a sense, our ....

R. J. Brachman. What IS-A is and isn't: An analysis of taxonomic links in semantic networks. IEEE Computer, 16:30--36, 1983.

....more rigorous and at the same time more understandable upperlevel ontologies. I will focus here on a single design practice, which is the main responsible of most semantic difficulties: IS A overloading. Notice that I do not refer with this expression to the old debate about the semantics of IS A [Brachman 1983], since I assume a standard set inclusion semantics for it. Rather, I will discuss the cases where this standard semantics does not reflect the ontological nature of the arguments. I will consider a number of bad practice cases, proposing a way to simplify the domain s hierarchical structure by ....

Brachman, R. 1983. What IS-A Is and Isn't: An Analysis of Taxonomic Links in Semantic Networks. IEEE Computer, 16(10): 30-36.

....concepts and their analytic interrelations, while the assertions stored in the ABox state contingent facts about what is true of the world. The TBox and ABox each have their own language. These are called the terminological language and the assertional language, respectively. Brachman and Levesque [7, 4] argue that these two components should be separately designed and optimized for their respective tasks. In particular, they insist on a strict separation of the inferential mechanisms each employs, in that neither component can change or manipulate the information contained in the other. ....

.... knowledge representation researchers since the latter part of 1988 (primarily as MIT LCS TM 387 and 387b) Our concern in this paper is solely with the restricted language and restricted classification theses, which have continued to exert influence on the field since their first appearances in [4, 6, 7, 19]. Subsequent to these articles and the writing of our paper, however, work by a number of authors (including Brachman, Levesque, and their students) has been concerned with improving the utility of knowledge representation systems in some of the directions we urge in our discussion, for example, ....

R. J. Brachman. What IS-A is and isn't: An analysis of taxonomic links in semantic networks. IEEE Computer, 16(10):30--36, October 1983.

....or their instances may even be totally unrelated. The set of given class derivations and their semantics ultimately determine how type and set relationships interact within a given data model as will be shown in Section 4. The term is a relationship has been misused to mean many different things [4]. We can now define the is a relationship in terms of the two just defined class relationships. Definition 5 C1 is a C2 ( C1 C2 and C1 C2. Informally, we say that C1 is a C2 if (1) every member of C1 is an member of C2 (the subset relationship) and (2) every property defined for C2 is also ....

R. J. Brachman, "What IS-A is and isn't: An Analysis of Taxonomic Links in Semantic Networks," in IEEE Computer, pp. 30 -- 36, Oct. 83.

....[16, 18] Therefore each tentative characterisation of OOL with regard to OBKR is contradicted by some non typical system. On the other hand, in the last ten years, there has been fundamental work concerning the semantics attached to the notions of objects, classes, specialisation and so on [3, 2]. There are roughly three ways for establishing the semantics of object based systems: designing an operational semantics which models the behaviour of the system without referring to the modelled domain [21] designing a denotational semantics given by an interpretation function ranging from ....

Ronald Brachman, What is-a is and isn't: an analysis of taxonomic link in semantic networks, IEEE Computer 16(10):30-36, 1983

....does not lie solely in its name [13] but rather in its constraint satisfaction and inheritance mechanisms. In this light, our ownership relationship can be viewed as a modeling primitive of an OODB system with built in semantics. In previous research, the IS A (or SUBCLASS) relationship [2, 12] and the PART OF relationship [6, 7, 8, 10, 11] have been modeled as semantic relationships. The remainder of this paper is organized as follows. In Section 2, we discuss the legal definition of ownership. In Section 3, we formally define the ownership relationship and describe its characteristic ....

R. J. Brachman. What IS-A is and isn't: An analysis of taxonomic links in semantic networks. Computer, 16(10):30--36, Oct. 1983.

....class, for example the class PERSON, will share common properties, such as name, home address, age, and social security number. We will also assume the existence of subclasses. Most semantic and object oriented databases use an IS A construct to support the concept of class and subclass. Brachman [Bra83] correctly notes that inheritance as defined by the IS A construct is really little more than convenient syntactic shorthand for incrementally creating subclasses, so we will ignore actual inheritance mechanisms per se. The important feature that is abstracted in the entity database model is the ....

R. J. Brachman, What IS-A Is and Isn't: An Analysis of Taxonomic Links in Semantic Networks, COMPUTER 16,10 (Oct. 1983), 30-36.

....: 17 4.1 Characterization of encoding schemes in terms of spanning set of down sets : 54 6.1 Asymptotic encoding results for theoretical orders : 77 6. 2 Empirical results (in bits) for chess learning system [16] : 78 6.3 Empirical results (in bits) for medical ontology : 78 xii List of Figures 1.1 Research overview : ....

....we not only gain the improved flexibility of sparse terms over bit vectors, but this shows that even for moderate size taxonomies, the asymptotic advantage of sparse terms pays off. CHAPTER 6. ENCODING WITH SPARSE LOGICAL TERMS 78 Table 6. 2: Empirical results (in bits) for chess learning system [16] Top Down Top Down Bottom Up Bottom Up Trans. Closure Compact Trans. Closure Compact Bit Vectors total 3,294,225 1,070,850 3,294,225 2,586,375 bits code 1,815 590 1,815 1,425 Sparse Terms total 820,872 803,056 966,920 1,007,104 bits code 452 442 533 555 Sparse Term Bit Vector ratio 0.25 0.75 0.29 ....

[Article contains additional citation context not shown here]

R. J. Brachman. What IS-A is and isn't: An analysis of taxonomic links in semantic networks. IEEE Computer, 16:30--36, 1983. BIBLIOGRAPHY 125

.... that may be either weak partial order relation (reflexive, transitive, antisymmetric) or strict partial order relation (irreflexive, transitive, antisymmetric) Now we assume that C rel covers weak partial order containment relations, for example concept subsumption [Woods, 1991] IS A relation [Brachman, 1983] and intensional containment [Kauppi, 1967; Palom ki, 1994] There are also part whole relations [Winston et al., 1987; Padgham and Lambrix, 1994; Sattler, 1995; Artale et al. 1996] that are either weak partial order relations or strict partial order relations. We assume that C rel can also be ....

Ronald J. Brachman, What IS-A is and isn't: An analysis of taxonomic links in semantic networks, IEEE Computer. Vol. 16, No. 10, 1983. Pp. 30-36.

....If every X is a Y , then the class of Xs inherits from the class of Y s. Brachman, however, distinguishes six different ways in which the is a relationship is actually used as a relation between classes and four different ways in which it is used as a relation between classes and individuals [6]. Although Brachman was writing about knowledge representation, his reasoning applies to object oriented programming as well. There are three inter class relations that most authors seem to agree upon: generalization, aggregation, and association. Roughly, generalization corresponds to is a and ....

R. Brachman. What IS-A is and isn't: an analysis of taxonomic links in semantic networks. IEEE Computer, 30--36, October 1983.

....EVENTS 2.1 Abstraction and Models Looking at the term abstraction one finds that there are a lot of definitions for it. It can be defined as the mental process of isolating a common element or explicating a relationship possessedby a number of things (Encyclop dia Britannica 1985) According to Brachman (1983) abstraction is a relation of type is a wherein a generic type is abstracted into an individual (e.g. the eagle in the eagle is an endangered species ) In (Encyclopedia of Artificial Intelligence 1985) the idea of abstraction is defined in the context of search as to at first ignore the ....

Brachman, R. J., 1983. What is-a is and isn't: an analysis of taxonomic links in semantic networks. Computer IEEE 16(10), pp. 30--36.

.... general that) another [2] These formalisms generally descend from the ideas presented in KL ONE [3] Term subsumption languages are more principled than both semantic networks and frames because the languages have welldefined semantics, which is often missing from frames and semantic networks [4, 5]. In the last few years many knowledge representation systems have been built using term subsumption languages, including KRYPTON, KL TWO, NIKL, BACK, SB ONE, LOOM, and CLASSIC. We will use the following example to illustrate definitional knowledge that can be expressed in term subsumption ....

R. J. Brachman, "What is-a is and isn't: An analysis of taxonomic links in semantic networks," Computer, vol. 16, no. 10, pp. 30--36, October 1983.

....involves the ability to form the kind of theory appropriate to the world knowledge of a system and understand the implications of the theory . The formal distinction between terminological knowledge and assertional knowledge is fundamental to work on the systems nikl (Schmolze [1989b] and krypton (Brachman et al. [1983, 1985] This work started in 1982. The main goal of the nikl project was to develop an enhanced terminological representation formalism, while the main goal of the krypton project was to incorporate a terminological representation formalism and a separate assertional representation formalism in ....

....In another attempt to clean up kl one, the unifying approach (Brachman and Levesque [1982] that combines a terminological representation formalism with an assertional representation formalism, was adopted in a host of so called hybrid knowledge representation systems. These include krypton (Brachman et al. [1983, 1985] kltwo (Vilain [1985] kandor (Patel Schneider [1984] meson (Edelmann and Owsnicki [1986] and back (Nebel and von Luck [1988] Nebel and von Luck [1988] define a hybrid knowledge representation formalism to consist of two or more different subformalisms for representing different ....

[Article contains additional citation context not shown here]

Brachman, R.J. [1983]. What IS-A Is and Isn't: An Analysis of Taxonomic Links in Semantic Networks. IEEE Computer 16 (10), 30--36.

....and make the message passing paradigm pervasive in the language. 2.4.2 Classic Meta Individuals. Classic [Brachman, et al. 1991] a modern descendent of KL ONE [Brachman and Schmolze, 1985] employs an approach to representing links on classes that has been called the abstraction relationship [Brachman, 1983]. This approach involves creating, as part of the language, a special kind of instance for each class which represents the class as an object. While these instances (called meta individuals) behave like all other instances (having links and being instances of a concept called concept) they have ....

Brachman, R. What IS-A is and Isn't: An Analysis of Taxonomic Links in Semantic Networks. IEEE Computer. 16(10). Pp. 30-36. Oct, 1983.

....for the particular fields of applications. This can be seen in distributed computing with the emphasis on mechanisms based on abstract data typing and subtyping. The same trend can also be witnessed in, for example, the AI community with the interest in more general forms of semantic relationships [29]. This development however challenges the very basis of object oriented computing. In particular, it becomes very difficult to define precisely what is meant by object oriented computing. It is clearly no longer possible to equate object orientation directly with the concepts of object, class and ....

Brachman, R.J., "What IS-A Is and Isn't: An Analysis of Taxonomic Links in Semantic Networks", IEEE Computer, pp 30-36, 1983.

....different, our definition of the ISA relationship is rather imprecise. It would be a worthwhile research endeavor to develop a more precise definition. 10] The artificial intelligence community has been struggling to precisely define ISA for over a decade. Rigorous studies of ISA can be found in [1, 2, 8, 19] (as well as the Etherington, Reiter and Hayes appears mentioned below) Brachman s KL ONE inference engine used its object hierarchies for its classification algorithm. KL ONE matches a specified concept and studies the known frames looking for known concepts that subsume the new concept. If ....

....as a subsumption of existing concepts. KL ONE s reliance on its taxonomy for its inferencing forced Brachman to precisely define ISA. Brachman catalogues the various usages of the ISA link seen in actual systems such as set membership, constraints for an individual, superset subset, containers [1]. After presenting the catalogue, Brachman makes the following surprising observation: One important observation to be made about our analysis of the ISA link is that inheritance of properties has played no part in our understanding. the useful (semantic) properties of ISA are. not pass this ....

[Article contains additional citation context not shown here]

R. Brachman. What is-a is and isn't: An analysis of taxonomic links in semantic networks. The AI Magazine, pages 66--73, October 1983.

....airport( BWI ) with the clause hairport(X) Gamma BWI unifies X.i. The clausal form is well suited for predicate rewrites in a variable substituted query. Let us call these clauses representing such a type taxonomy 4 isa clauses after the ISA links found in the semantic networks literature [Bra83] In the future, we could easily operate in an alternative language domain in which types are handled by a separate means, as in LOGIN [AKN86] or use the notational conveniences in [MM77] In truth, any first order, Horn clause theory describes a taxonomy of concepts, and represents a semantic ....

Ronald J. Brachman. What IS-A Is and Isn't: An Analysis of Taxonomic Links in Semantic Networks. IEEE Computer, pages 30--36, 1983.

.... network) belongs to (in Aristotle s syllogism) subset (in set theory) inheritance assertion (in inheritance network [35] as well as many relations studied in psychology and philosophy, such as type token , category instance , general specific , and superordinate subordinate [7]. What make it different from the others are: it is a relation between two terms, and the 4 relation is completely defined by the two properties: reflexivity and transitivity. 1 This logic (as well as the following two) can be interpreted in the usual model theoretic way. Terms have no ....

.... the four types of extensional inheritance can be represented in Figure 1 [4] where there are four types of relations: 3 Here inheritance is used for a logical relation between two terms, rather than an idea about the implementation of a knowledge base, by which storage space can be saved [7, 35]. 7 1. If S a P is true, then S i P is true; if S e P is true, then S o P is true. 2. S a P and S o P must be one true and one false; S e P and S i P must be one true and one false. 3. S a P and S e P cannot both be true. 4. S i P and S ....

R. Brachman. What is-a is and isn't: an analysis of taxonomic links in semantic networks. IEEE Computer, 16:30--36, 1983.

....from it. Results for a multi scale, multi abstraction level road extraction system, which is based on these considerations, are presented, and conclusions are given. 2. Abstraction and Scale Space Events 2.1. Abstraction and Models There are a lot of definitions for the term abstraction (e.g. [6]) In this paper a special notion of abstraction is used. It is defined in the context of image understanding where symbols are mapped to portions of images. The description by means of the symbols has to be structured. Additionally it has to be simplified, emphasis has to be laid on important ....

R.J. Brachman. What IS-A Is and Isn't: An Analysis of Taxonomic Links in Semantic Networks. IEEE Computer, 16(10):30--36, 1983.

.... generalization, i.e. commonalities and differences of objects are organized by answering the question: Which properties of objects are more general and which are more specific This view of inheritance is not taken as a matter of course [Str88] The old debate about what inheritance is or isn t [Bra83] is still alive. Rooting in AI, especially in knowledge representation, the metaphor of inheritance has been approved for standard programming languages, e.g. C . Many insights are going to be reinvented by a different community from a different point of view. So far, more attention has been ....

Ronald J. Brachman. What IS-A Is and Isn't: An Analysis of Taxonomic Links in Semantic Networks. IEEE Computer, 16(10):30--36, October 1983.

....be specified by using two kinds of abstractions: aggregation and generalisation [Smith77] Abstraction of aggregation is based on the presented set of binary dependencies and special aggregation operations among concepts. The basis for a generalisation abstraction constitute the inheritance (ISA [Brach83]) dependency. The inheritance dependency can be described in terms of binary dependencies. Several generalisation constraints, which are identified in the area of semantic data modelling [HuKi87] PeMa88] can be defined by using inheritance dependency, and two intensional(sum and product ....

....with basic dependencies allows us to describe the semantics of more complicated constructions. For example, it is possible to define the semantics such categories as time stamped entity and time stamped relationship [Tempora92] 4.2. 2 Abstraction of Generalisation Inheritance (ISA) dependencies [Brach83], HuKi87] PeMa88] are of various kinds. Although the inheritance constructs are well understood [GuOr94] there is no complete agreement of interplay between the inheritance dependency and other types of basic constraints. For example, some of the researchers understand the inheritance in the ....

R Brachman. What IS-A is and isn't: An analysis of taxonomic links in semantic networks. Computer,16(10), pp. 30-36, 1983.

....by assigning them to a class. All entities in the class, for example the class PERSON, will share common properties, such as name, home address, age, and social security number. Most semantic and objectoriented databases support the concept of sub class by means of an IS A construct. Brachman [Bra83] correctly notes that inheritance as defined by the IS A construct is really little more than convenient syntactic shorthand for incrementally creating sub classes, so we will ignore actual inheritance mechanisms per se. The important feature we abstract in the entity database model is the ....

....C i = F i , E i ) is said to be a sub class of C k = F k , E k ) denoted C i C k , if (a) F i F k , and (b) E i E k (that is, E i logically implies E k ) Condition (a) seems to be universally accepted in both object oriented and semantic network class hierarchies. Several authors (c.f. [Bra83, Tou86]) have found strict implication too strong for many forms of semantic reasoning. They would replace it with a non monotonic logic that allows exceptions. Nevertheless, a condition similar to (b) would seem to be essential in sub4 class definition. Given this definition, it is easy to show that ....

R. J. Brachman, What IS-A Is and Isn't: An Analysis of Taxonomic Links in Semantic Networks, COMPUTER 16,10 (Oct. 1983), 30-36.

....the specification and to the implementation of these procedures. The procedure given in [16] uses a syntax based on taxonomic relations between expressions that denote sets. This procedure can be viewed as an extension of earlier work on knowledge representation languages, e.g. 3] 11] 4] [5]. Knowledge representation languages have traditionally been organized around taxonomic relationships between classes. AI researchers often express the intuition that such taxonomic representations facilitate inference. The technical results in [16] support this intuition. The relationship between ....

R. J. Brachman. What is-a is and isn't: An analysis of taxonomic links in semantic networks. IEEE Computer, 16(10):30--36, October 1983.

....in particular by using the Constraint language (see later) The fact that the classes C 1 and C 2 are related by the ISA relation means that every instance of C 1 is also an instance of C 2 . The importance of the ISA relation in representing knowledge is stressed in many papers (see, for example, [9]) and stems from the fact that it allows a modular approach to schema design, based on the inheritance of properties. Formally, the Typed Graph is a labeled graph, where the set of nodes is partitioned into two sets, one representing classes, and the other one representing roles. The arcs ....

R.J. Brachman, "What IS-A is and isn't: An Analysis of Taxonomic Links in Semantic Networks," IEEE Computer, Vol. 16, N.10, pp. 30-36, 1983.

....airport( BWI ) with the clause hairport(X) Gamma BWI unifies X.i. The clausal form is well suited for predicate rewrites in a variable substituted query. Let us call these clauses representing such a type taxonomy 4 isa clauses after the ISA links found in the semantic networks literature [Bra83] In the future, we could easily operate in an alternative language domain in which types are handled by a separate means, as in LOGIN [AKN86] or use the notational conveniences in [MM77] In truth, any first order, Horn clause theory describes a taxonomy of concepts, and represents a semantic ....

R. J. Brachman. What IS-A Is and Isn't: An Analysis of Taxonomic Links in Semantic Networks. Computer, 16(10):30--36, October 1983.

No context found.

R. J. Brachman. What IS-A is and isn't: An analysis of taxonomic links in semantic networks. IEEE Computer, 16(10):30--36, 1983.

No context found.

R. J. Brachman, "What IS-A Is and Isn't: An Analysis of Taxonomic Links in Semantic Networks", IEEE Computer, vol. 16, pp. 30-36, 1983.

No context found.

R.J. Brachman. What IS-A is and isn't: An analysis of taxonomic links in semantic networks. IEEE Computer, 16(10), 1983.

No context found.

Brachman RJ. What IS-A is and isn't: an analysis of taxonomic links in semantic networks. Computer 1983;16(10):30-6.

No context found.

R.J. Brachman. What IS-A Is and Isn't: An Analysis of Taxonomic Links in Semantics Networks. IEEE Computer, 16(10):30-36, 1983.

No context found.

Ronald J. Brachman. What is-a is and isn't: An analysis of taxonomic links in semantic networks. Computer, 16(10):30--36, October 1983.

No context found.

Brachman, Ronald J. 1983. What IS-A is and isn't: An analysis of taxonomic links in semantic networks. IEEE Computer, 30-36.

No context found.

R.J. Brachman (1983) What IS-A is and isn't: An Analysis of Taxonomic Links in Semantic Networks.<F3.17e+05> IEEE Computer<F4.675e+05><F3.733e+05> 16: 30--36,

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