Brachman, Ronald J.; 1985. I lied about the trees. AI Magazine 6(3).  Home/Search   Document Not in Database   Summary   Related Articles   Check

....to the class object as a whole, not applicable to its subclasses or instances, in contrast to instance attributes. Use of class attributes is not discussed in this paper. Introduction of the distinction between essential and non essential attributes is motivated by the issues addressed in [15], concerning the definitional inability of object based formalisms. One of the Brachman s points in [15] is that, due to the fact that most of the object based formalisms do not distinguish between cancellable (incidental) and uncancellable (essential) properties, one can in principle create an ....

....attributes. Use of class attributes is not discussed in this paper. Introduction of the distinction between essential and non essential attributes is motivated by the issues addressed in [15] concerning the definitional inability of object based formalisms. One of the Brachman s points in [15] is that, due to the fact that most of the object based formalisms do not distinguish between cancellable (incidental) and uncancellable (essential) properties, one can in principle create an instance of a class that overrides all the properties stored in the class. This is not allowed in SILO. ....

R. J. Brachman, "I Lied about the Trees" Or, Defaults and Definitions in Knowledge Representation, AI Magazine Fall 85 (1985) 80-93.

....isn t known to be a penguin then the KQL interpreter will conclude (if asked) that Tweety flies. If later it is asserted that Tweety is a penguin after all, it will not undo the flies assertion 1 . This mixture of defeasible assertions with monotonic inference may seem like trouble (e.g. see [3]) but in practice trouble has rarely arisen. 3.1.2 Defeasible Reasoning: Drawing Tentative Conclusions 3. Typically, liquids are applied with a brush. 4. Typically, leaves are green. KM, KQL and Algernon are all monotonic systems they only compute the consequences logically implied 2 ....

I. Brachman. `i lied about the trees', or, defaults and definitions in knowledge representation. AI Magazine, 6(3), 1985.

....in certain cases. It can be argued that these anomalies are a result of using the same inference procedure on assertions which have different meanings. That is, it would seem that these systems can deal with certain types of typicality assertions but not with other types. For example, Brachman [13] has pointed out that there is a difference between prototypical properties, which are characteristic of a kind, and properties which typically apply to instances of a kind. For example, birds lay eggs is a prototypical property of birds, but we would not want to assume by default that a given ....

Ronald J. Brachman. I lied about the trees. AI Magazine, 6(3):80--93, 1985.

....interaction a key aspect of computation that is not handled by the traditional definition [Wegner, 1997] We support the intuition that interaction is crucial, and seek to incorporate it below in our treatment of agenthood. Tests. It is widely recognized that definitions can never be perfect [Brachman, 1985] . This is because definitions are meant to provide necessary and sufficient conditions, but they usually provide neither because of exceptions. The concept of agents in general appears to be subjective and not crisp. Consequently, and especially given the early stage of agents research, we ....

Brachman, Ronald J.; 1985. "I lied about the trees", or, defaults and definitions in knowledge representation. AI Magazine 6(3):80--93.

....Alternatively, one can view inheritance as making a public declaration that objects of the child class obey the semantics of the parent class, so that the child class is merely specializing or refining the parent class. This question is addressed in the context of knowledge representation in [Brachman85]. Our position is that being able to use inheritance without making a public commitment to it in the external interface of a class is valuable, and we will analyze how existing object oriented languages support this option. We will begin by considering only the single inheritance case, i.e. where ....

Ronald J. Brachman. "I Lied about The Trees" Or Defaults and Definition in Knowledge Representation. AI Magazine 6:3 (Fall 1985), 80-93.

....controlled by the obkrs. Among other conditions, the result must belong to the domain of the attribute the local inference mechanism is attached to. Relevant information about filters may be found in [Dek94] Win85, Rec93] contain information about procedural 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 ....

R.J. Brachman. "i lied about the trees", or defaults and definitions in knowledge representation. AI Magazine, 6(3):80--93, 1985.

....language of inheritance. Introduction 4 definitions of path preference in an inheritance system as that system s proof theory . Unfortunately, however, a semantic theory which unifies the various approaches to the proof theory of inheritance reasoning has yet to be developed (Brachman, 1983, Brachman, 1985, Touretzky, et al. 1987, Boutilier, 1989, Dorosh and Loui, 1989, Delgrande, 1990) Links in an inheritance network notation can be used to represent the natural language qualification, typically as in the sentence, Typically, humans have two legs (cf. Network AB typically contains ....

....nets, and the two sorts of links are treated as fundamentally different. The combination of different sorts requires different rules. The heterogeneous system addresses Brachman s criticisms that defeasible nets Path Based Inheritance 35 cannot represent definitions of composite concepts (cf. Brachman, 1983, 1985)) Horty and Thomason (1988) have also addressed this criticism. A link or ER can be positive or negative, strict or defeasible. Rules govern the computation of inclination qualifiers associated with various combinations. Some combinations, like the chaining of any link after a negative (strict or ....

Brachman, Ronald J. (1985). I Lied about the Trees, or Defaults and Definitions in Knowledge Representation. The AI Magazine, 6, 80-93.

....the ability to quantify over concept definitions. 19 Synonyms and abbreviations constitute trivial but useful forms of multiple definitions. Doyle Patil central element of taxonomic inference, mainly because there seem to be very few strictly necessary conditions in most natural kinds (cf. [5, 16]) Instead of being definable by sets of necessary and sufficient conditions, natural kinds appear to be conceptually convenient bundles of default properties which describe the prototype to which their instances bear family resemblances. Though the prevalence of natural kinds and the importance ....

R. J. Brachman. `I lied about the trees,' or, Defaults and definitions in knowledge representation. AI Magazine, 6(3):80--93, 1985. Doyle & Patil

....15 Classifying FRSs do not provide inheritable default values that can be canceled because the notion of cancellation conflicts with the notion of necessary conditions. That is, classification implies the comparison of definitions, and definitions are true by necessity, not by default [10]. The notion of definition also implies that a child frame cannot inherit conflicting information from its parents since conflicting definitions would indicate an inconsistency in a KB. FRSs that classify do not provide metaclasses, seeUnits, facets, nor own slots and member slots. Classifying ....

R.J. Brachman. I lied about the trees. AI Magazine, 6(3):80--93, 1985.

....of this diversity so that the concepts can be compared. Our point of view is structural, but some of the behavioral aspects are also considered. For an excellent treatment of the behavioral aspects of inheritance in programming languages see [WE89] For inheritance in artificial intelligence see [4]. Most of the examples will use C [6] because this is the most common programming language supported by commercial object oriented database management systems. We begin with a list of some of the variations on the general theme of inheritance. In this list, the concepts are merely introduced. We ....

R. Brachman, "I lied about the trees" or, defaults and definitions in knowledge representation, AI Magazine, pages 80--93, Fall, 1985.

....definitions, for instance, determining subsumption relationships between concepts and instance relationships between objects and concepts two kinds of inferences we will collectively refer to as classification. Although these two forms of representation and reasoning may seem to be incompatible [2], it would of course be desirable to combine them. From the point of view of nonmonotonic inheritance networks, it would be interesting to have a richer description language for specifying classes and properties and to add the ability of classifying objects as belonging to some class. From the ....

....depicted by a letter, we also allow for defined nodes, depicted by an encircled letter. The latter nodes are assumed to be defined by the conjunction of all nodes that are reachable by a single strict positive link. As an example let us consider the following small knowledge base (inspired by [2]) Elephant v Mammal=legs: 4; color: grey Hepatitis Elephant : Elephant u infected by: Hepatitis=color: yellow Yellow Elephant : Elephant u color: yellow x: Elephant infected by(x) Hepatitis Using the abbreviations M , E, H, and Y for Mammal, Elephant, Hepatitis Elephant, and ....

R. J. Brachman. `I lied about the trees' or, defaults and definitions in knowledge representation. The AI Magazine, 6(3):80--93, 1985.

....Sauce Make Spaghetti Pesto Make Spaghetti Marinara Make Chicken Marinara Make Meat Dish Make Fettuccine Make Alfredo Sauce Make Pesto Make Spaghetti Make Marinara Fig. 1. An explanatory hierarchy in the cooking world. 4 The abstraction axioms can be seen as ISA relationships without exceptions [1]. Even if such axioms are not syntactically different from the explanatory ones, it is important to maintain them distinct. Figure 1 presents an example of explanatory hierarchy adapted from [14] thick grey arrows correspond to abstraction axioms, as in [14] while thin black arrows correspond ....

R. Brachman. "I lied about the trees" Or defaults and definitions in knowledge representation. AI Magazine, 6:80--93, 1985.

....and roles are analogous to frames and slots, respectively. Since its basic structural units are concepts, Brachman and Schmolze [1985] describe kl one as a concept oriented formalism. Unlike frames, klone concepts allow neither defaults nor user intervention through attached procedures. Brachman [1985] shows why defaults and attached procedures are not compatible with kl one based knowledge representation. Note that the earliest versions of kl one did in fact allow attached procedures see Woods and Schmolze [1991] Brachman and Schmolze [1985, p. 179] described concepts as (frame like) ....

....to be unsolvable in polynomial time. However, subsumption in FL Gamma , a variant of FL that includes all the operators of FL but one (the role restriction operator) has quadratic time complexity and is tractable. As a consequence the role restriction operator was omitted in the later version (Brachman et al. [1985]) of krypton. A small increase in the expressiveness in FL Gamma to FL, therefore, results in a dramatic increase of the computational complexity, from tractable to intractable. The reader interested in the theory of computational complexity and undecidability is advised to refer to an ....

Brachman, R.J. [1985]. "I Lied about the Trees" or, Defaults and Definitions in Knowledge Representation. AI Magazine 6 (3), 80--93.

....class can be defined as a subclass of the empty class, which we will call Object. This is actually enforced in Smalltalk. In this situation, the inheritance hierarchy is a tree, otherwise it is a forest. If C is the superclass and D is the subclass, then it is common to say that D is a C [6, 55]. For example, if Student is a subclass of Person, then it seems reasonable to say that Student is a Person. It is convenient to let is a denote the transitive closure of this relation. The other possible relation between classes is has a. If a class C declares a variable of a type which contains ....

Ronald J. Brachman. "I lied about the trees" or, defaults and definitions in knowledge representation. The AI Magazine, 6(3):80--93, 1985.

....this can always be done. Simply specify x by cancelling all properties of y that are not true of x and adding all properties of x that y does not exhibit. This point is neatly illustrated by a delightful anecdotal example due to Brachman, consisting of a question and a series of answers [Brachman, 1985]: Q: What s big and gray, has a trunk, and lives in the trees A1: An elephant, I lied about the trees. A2: A giraffe I lied about the color, the trunk and the trees. A3: An idea I lied about the color, the trunk, the trees, and about the lives . 1 However, some recent description ....

....They concluded that designers of description logic systems must make careful choices in trading expressiveness for tractability. Moreover, there is no single best choice. Instead, different choices may be appropriate depending on the application. 7 Much of which is included in [ Levesque and Brachman, 1985 ] 30 The result of Brachman and Levesque has practical significance because their co NP complete language is a subset of the languages employed by such systems as kl one. Nebel later showed that another subset of the languages used in systems like kl one and back creates an NP hard ....

R. J. Brachman. "i lied about the trees" (or, defaults and definitions in a knowledge representation). AI Magazine, 6(3):80--93, 1985.

.... Inheritance Networks proposed by Brachman and realized in the kl one system (Brachman, 1979) This novel paradigm was motivated by the shortcomings of semantic networks, which have been criticized for their lack of formal semantics which results in ambiguity and contradictions (Woods, 1975; Brachman, 1985). During the last ten years, this line of research has led to the development of various TKRSs (see for example (Woods and Schmolze, 1992; SIGART, 1991; Baader et al. 1994) These systems make a clear distinction between intensional knowledge, or general knowledge about the problem domain, and ....

Brachman, R. J. (1985). "I lied about the trees". AI Magazine, 6(3):80--93.

....what can be inferred. 4.1 Specificity for Conflict Resolution A major thrust of work on direct argumentation systems has been to derive general principles for adjudicating among conflicting lines of argumentation. The most prominent method is that of taxonomic inheritance with exceptions (e.g. [7, 1]) Here, the principle of specificity is used to resolve conflict: information about subclasses overrides information about more general classes in the case of conflict. If Tweety is a bird, assume he flies, but if Tweety is an ostrich then assume he does not. However, while intuitively appealing ....

I. Brachman. `i lied about the trees', or, defaults and definitions in knowledge representation. AI Magazine, 6(3), 1985.

....of Inference This chapter introduces the basic notation and proof theoretic considerations at stake for inheritance reasoning. Default inheritance has been proposed as a (sometimes; see below) tractable alternative to first order logic as a model of human reasoning with natural language generics. Brachman s (1985) arguments about the impossibility of representing analytic truths in a purely default framework noted, there is reason to think that nearly all knowledge is expressed in sentences with respect to which exceptions exist. Inheritance logics make typicality a primitive and therefore offer ....

Brachman, R. J. (1985). I Lied about the Trees, or Defaults and Definitions in Knowledge Representation. The AI Magazine, 6, 80--93.

....as in [ Kautz, 1991 ] that there is knowledge about entities at different levels of abstraction, including properties (such as causes and effects) of the entities and hierarchical dependencies among the entities. Abstraction axioms, which can be seen as IS A relationships without exceptions [ Brachman, 1985 ] are a direct way of introducing abstractions in the logical approach to abduction. They are sufficient for discussing the declarative and computational issues mentioned at the beginning of this chapter, even if they do not necessarily correspond to all of the ways abstraction can be used or ....

Brachman, R. 1985. "I lied about the trees" Or defaults and definitions in knowledge representation. AI Magazine 6:80--93.

....use terminological systems for this kind of commonsense reasoning, one needs a formalism that can handle such default assumptions, but does not destroy the definitional character of concept descriptions because otherwise the advantage of automatic concept classification, etc. would be lost (see [5]) Besides the general arguments for the importance of reasoning with defaults, which can be found in the nonmonotonic reasoning literature, the need for embedding defaults into terminological representation formalisms is also substantiated by the fact that this is an important item on the wish ....

R. J. Brachman. `I lied about the trees' or, defaults and definitions in knowledge representation. The AI Magazine, 6(3):80--93, 1985.

....documentation, to guide both end users and any future maintainers of the program. Standards for documenting knowledge systems have been slow to emerge. This has often led to worries about assumptions implicit in the system. For example, the discussions about the interpretation of is a links (e.g. Brachman, 1986)) in which it can be unclear as to whether they are intended to be exhaustive, exclusive, and whether they are used consistently. An ontology provides much of the documentation which is required and provides definitive answers to questions of this sort. 4) Knowledge acquisition. Too often ....

Brachman, R.J. (1986). I Lied About the Trees, AI Magazine, Vol.6, No.3, pp.60-93.

....was to integrate reason maintenance, constraints and objects into a well defined system, with a precisely defined semantics. Many objections have been made to frame systems. The lack of a clear semantics, especially for inheritance, among them. These drawbacks are usually [Brachman et al. 1983, Brachman, 1985 ] inspired by a strong assumption of what knowledge representation is. Namely, that a representation is possible that does not depend upon the scope. Our purpose is more restricted. On one hand, we go back to the original motivation of frames, which was to separate the description of world into ....

R. J. Brachman. "I lied about the trees", or, defaults and definitions in knowledge representation. The AI Magazine, pages 80--93, Fall 1985.

.... roots in Structured Inheritance Networks proposed by Brachman and realized in the kl one system [Bra79] This novel paradigm was motivated by the shortcomings of semantic networks, which have been criticized for their lack of formal semantics which results in ambiguity and contradictions [Woo75, Bra85] During the last ten years, this line of research has led to the development of various TKRSs (see for example [WS92, SIG91, BFH 94] These systems make a clear distinction between intensional knowledge, or general knowledge about the problem domain, and extensional knowledge, which is ....

Ronald J. Brachman. "I lied about the trees". AI Magazine, 6(3):80--93, 1985.

No context found.

Brachman, Ronald J.; 1985. I lied about the trees. AI Magazine 6(3).

No context found.

Brachman, Ronald J.; 1985. I lied about the trees. AI Magazine 6(3).

No context found.

Brachman, Ronald J. (1985) "`I Lied About the Trees", Or, Defaults andDefinitions in Knowledge Representation.' The AI Magazine, 6.3, Fall : pp. 80--93.

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