Showing papers on "Knowledge acquisition published in 1991"

Knowledge Discovery in Databases

Abstract: From the Publisher: Knowledge Discovery in Databases brings together current research on the exciting problem of discovering useful and interesting knowledge in databases. It spans many different approaches to discovery, including inductive learning, bayesian statistics, semantic query optimization, knowledge acquisition for expert systems, information theory, and fuzzy 1 sets. The rapid growth in the number and size of databases creates a need for tools and techniques for intelligent data understanding. Relationships and patterns in data may enable a manufacturer to discover the cause of a persistent disk failure or the reason for consumer complaints. But today's databases hide their secrets beneath a cover of overwhelming detail. The task of uncovering these secrets is called "discovery in databases." This loosely defined subfield of machine learning is concerned with discovery from large amounts of possible uncertain data. Its techniques range from statistics to the use of domain knowledge to control search. Following an overview of knowledge discovery in databases, thirty technical chapters are grouped in seven parts which cover discovery of quantitative laws, discovery of qualitative laws, using knowledge in discovery, data summarization, domain specific discovery methods, integrated and multi-paradigm systems, and methodology and application issues. An important thread running through the collection is reliance on domain knowledge, starting with general methods and progressing to specialized methods where domain knowledge is built in. Gregory Piatetski-Shapiro is Senior Member of Technical Staff and Principal Investigator of the Knowledge Discovery Project at GTELaboratories. William Frawley is Principal Member of Technical Staff at GTE and Principal Investigator of the Learning in Expert Domains Project.

Probabilistic Similarity Networks

Abstract: I address practical issues concerning the construction of normative expert systems--expert systems that encode knowledge within a decision-theoretic framework. In particular, I examine the similarity network and partition, two extensions to the influence diagram. A similarity network is a tool for building an influence diagram, whereas a partition is a tool for assessing the probabilities associated with an influence diagram. Both representations encode asymmetric forms of conditional independence that are not represented conveniently in an ordinary influence diagram. Similarity networks and partitions exploit these forms of conditional independence to facilitate the construction and assessment of influence diagrams for problems of diagnosis. The representations aided considerably the construction of Pathfinder, a large normative expert system for the diagnosis of lymph-node diseases (the domain contains approximately 60 diseases and 110 disease findings). In an early version of the system, I encoded the knowledge of the expert using an erroneous assumption that all disease findings were conditionally independent, given each disease. When the expert and I attempted to build an influence diagram for the domain to capture the dependencies among the disease findings, we failed. Using a similarity network, however, we were able to construct the influence diagram for the entire domain in approximately 40 hours. Furthermore, using the partition representation, the expert was able to decrease the time required to assess a probability--on average--by almost one order of magnitude. Most important, through a comparison procedure based in decision theory, I found that the improvements in diagnostic accuracy afforded by the more sophisticated model of the domain were well worth the additional effort that we had invested to build the revised version of the system. In this work, I examine in detail the theoretical properties of similarity networks and partitions, and discuss the application of these representations to the construction of Pathfinder. This research suggests strongly that, by identifying specific forms of conditional independence, and by developing representations that exploit these forms of independence for knowledge acquisition, knowledge engineers can construct normative expert systems for domains of larger scope and greater complexity than the domains previously through to be amenable to the decision-theoretic approach.

The Requirements Apprentice: automated assistance for requirements acquisition

Abstract: An automated tool called the Requirements Apprentice (RA) which assists a human analyst in the creation and modification of software requirements is presented. Unlike most other requirements analysis tools, which start from a formal description language, the focus of the RA is on the transition between informal and formal specifications. The RA supports the earliest phases of creating a requirement, in which ambiguity, contradiction, and incompleteness are inevitable. From an artificial intelligence perspective, the central problem the RA faces is one of knowledge acquisition. The RA develops a coherent internal representation of a requirement from an initial set of disorganized imprecise statements. To do so, the RA relies on a variety of techniques, including dependency-directed reasoning, hybrid knowledge representations and the reuse of common forms (cliches). An annotated transcript showing an interaction with a working version of the RA is given. >

Knowledge acquisition for knowledge-based systems

Abstract: The work reported at the first Japanese Knowledge Acquisition for Knowledge-Based Systems Workshop is discussed, providing both an overview of the field and an introduction to a series of articles on knowledge acquisition. The discussion covers tools, methods, and mediating representations; real-time problem solving; the system-model-operator metaphor; an interview architecture based on dynamic analysis, inductive knowledge acquisition from structured data; research in Japan; how to make application programming easier; justification-based knowledge acquisition; integrating knowledge acquisition and performance systems; tasks, methods, and knowledge; rule induction; hypertext; explanation-based learning and case-based reasoning; and interviewing. >

The Use of Background Knowledge in Decision Tree Induction

Abstract: At present, algorithms of the ID3 family are not based on background knowledge. For that reason, most of the time they are neither logical nor understandable to experts. These algorithms cannot perform different types of generalization as others can do (Michalski, 1983s Kodratoff, 1983), nor can they can reduce the cost of classifications. The algorithm presented in this paper tries to generate more logical and understandable decision trees than those generated by ID3-like algorithmss it executes various types of generalization and at the same time reduces the classification cost by means of background knowledge. The background knowledge contains the ISA hierarchy and the measurement cost associated with each attribute. The user can define the degrees of economy and generalization. These data will influence directly the quantity of search that the algorithm must undertake. This algorithm, which is an attribute version of the EG2 method (Nunez, 1988a, 1988b), has been implemented and the results appear in this paper comparing them with other methods.

Intelligent Assistant Systems

Abstract: Representation and models - knowledge representation, general aspects, logic and objects, situational versus analytical knowledge symbolic reasoning - search, production systems, problem solving uncertainty and belief revision - representation of uncertainty, belief revision human-machine interaction - sharing intelligence, user interfaces, advanced interaction media, knowledge acquisition.

Designing curricula for conceptual restructuring: Lessons from the study of knowledge acquisition in astronomy

Abstract: This document examines children's and adults' knowledge of observational astronomy and characterizes the kinds of mental models students form when asked questions in astronomy. Mental models were grouped into three categories: intuitive, synthetic, and scientific. Implications for the design of curricula and for instruction are identified. In designing curricula in domains where learning requires the restructuring of prior knowledge, particular attention must be paid to the sequence in which the various concepts that comprise a given domain are introduced. It is suggested that instruction consistent with the sequence of acquisition of these concepts will be more successful than instruction that is not. The texts written should provide adequate explanations of the scientific concepts that are introduced, explanations that take into account the mental models and entrenched beliefs the students may have based on their everyday experience. Particular attention must be paid to providing students with situations that make them realize that what they may consider as facts about the world may be interpretations subject to falsification, and that sometimes there can be good reasons for replacing their existing beliefs with a new explanatory framework. Contains 35 references. (Author/MDH) *********************************************************************** Reproductions supplied by EDRS are the best that can be made from the original document. *********************************************************************** CENTER FOR THE STUDY OF READING

A Practical Guide to Knowledge Acquisition

Abstract: From the Publisher: This book presents knowledge acquisition as an integral part of building an expert system. It describes successful knowledge acquisition techniques relevant to all expert-systems applications. It discusses problems that anyone building such a system is likely to encounter, how to avoid these problems, and how to correct them when they do occur. Guidelines explain not only how to ask questions in interviews, but what questions to ask. 0201145979B04062001

Generating fuzzy rules by learning from examples

Abstract: A general method is developed for generating fuzzy rules from numerical data. The method consists of five steps: dividing the input and output spaces of the given numerical data into fuzzy regions; generating fuzzy rules from the given data; assigning a degree to each of the generated rules for the purpose of resolving conflicts among the generated rules; creating a combined fuzzy-associative-memory (FAM) bank based on both the generated rules and linguistic rules of human experts; and determining a mapping from input space to output space based on the combined FAM bank using a defuzzifying procedure. The mapping is proved to be capable of approximating any real continuous function on a compact set to arbitrary accuracy. The method is applied to predicting a chaotic time series. >

An Efficient Interlingua Translation System for Multi-lingual Document Production

Abstract: Knowledge-based interlingual machine translation systems produce semantically accurate translations, but typically require massive knowledge acquisition. This paper describes KANT, a system that reduces this requirement to produce practical, scalable, and accurate KBMT applications. First, the set of requirements is discussed, then the full KANT architecture is illustrated, and finally results from a fully implemented prototype are presented.

Machine learning : proceedings of the Eighth International Workshop (ML91)

Abstract: The proceedings of the Eighth International Workshop (ML91) held at Northwestern U., Evanston, Illinois, in June 1991. All papers contain new work, new results, or major extensions to prior work. Topics include automated knowledge acquisition, computational models of human learning, constructive ind

A resource-based paradigm for the configuring of technical systems from modular components

Abstract: In the resource-based paradigm, the interfaces through which technical systems, their components, and their environment interact are modeled as abstract resources, and each technical entity is characterized by the types and amounts of resources it supplies, consumes and uses. This intuitive model, derived in one application area, is shown to be in concordance with the design rationale of modular component systems. A simple self-organizing configuring inference procedure for the resource-based paradigm, resource-balancing, with a description of the environment of the technical system as the requirement specification, is derived from the basic acceptance criterion for configurations. Five levels of knowledge are defined for this paradigm and introduced in a simple representation scheme which, through its inherent locality and mutual isolation of component knowledge, allows efficient acquisition and maintenance of even large component knowledge bases. First experiences with the implementation and use of these ideals in the prototype shell COSMOS are reported. >

The KAOS Project: Knowledge Acquisition in Automated Specification of Software

Abstract: A dry semipermeable membrane, which is adapted to separate solutes from solutions by reverse osmosis methods, is prepared by a process which comprises casting a film from a casting solution containing a cellulose derivative, a first additive, and a suitable amount of an organic compound with an organic solvent as the balance of the casting solution; evaporating a portion of said organic solvent from said cast film; dipping the partially evaporated film in water to extract the remainder of said organic solvent and drying the extracted film.

Using a description classifier to enhance deductive inference

Abstract: Descriptions are given of some of the emerging techniques and uses of classifier-based reasoning systems, specifically as they apply to the LOOM knowledge representation system. The author asserts that current generation expert system tools fail to achieve a satisfactory integration of frame knowledge and rule knowledge. He then describes a class of languages, exemplified by LOOM, that combine descriptions and rules to form a hybrid logic that does achieve a satisfactory level of integration. The use of classifier technology enables a form of unification over descriptions that fills a gap present in the frame-plus-rule (F+R) technology. In addition, classification-based inference technology is more powerful than the inference technology found in languages such as (pure) Prolog. A classifier's ability to automatically organize definitions and to detect many kinds of inconsistency can significantly benefit the task of knowledge acquisition. The unique capabilities of the classifier can be applied to enhance existing programming paradigms. The author highlights specific enhancements to the production rule and object-oriented programming paradigms. >

Meaningful Learning in College Biology through Concept Mapping.

Abstract: A CCORDING to course descriptions and teachers' intensions, undergraduate science courses should help students develop critical thinking skills and increase their understanding of the technological and natural environment in which they live and work. In other words, as teachers, we hope that students will learn science concepts in a meaningful way and be able to apply those concepts to solving real problems. Yet, many college students would admit that they are unable to remember concepts past the final examination and do not connect science concepts to life experiences. Students rarely consider whether the knowledge they obtain will be lasting to them or meaningful in their lives. Ausubel (1968) suggested that individuals learn meaningfully by building knowledge on the basis of what they already know. More recently von Glasersfeld (1981; 1988) has suggested an epistemological theory that he terms radical constructivism. The constructivist view of knowledge acquisition holds that construction of knowledge is a personal activity in which the selection, interpretation and reorganization of sensory data varies depending on the individual's prior knowledge. Therefore the conceptualization of one individual can only be similar to another's, but cannot match it. Traditional approaches to teaching do not take into account the constructivist theory of knowledge acquisition. Faced with typical didactic classroom presentations and multiple choice assessments, students view their role in learning to be taking as much information as possible from the teacher or text and storing it by rote memorization. Passing the examination becomes the goal which drives study, and little thought is given to learning with understanding. If we are to facilitate meaningful learning for students, we must provide them with a purpose for rejecting rote learning strategies while simultaneously assisting them with developing strategies that lead to learning with understanding. In this article we report on the potential of concept mapping, a learning strategy, coupled with assessments oriented toward problem solving, to facilitate meaningful learning in a freshman biology class. Interpretive methods were used to answer two questions:

Instructional Conversations and their Classroom Applications

Abstract: Author(s): Goldenberg, Claude | Abstract: Generations of educators have advocated a type of teaching that does more than impart knowledge and teach skills. Knowledge and skills are undoubtedly important, but true education requires far more. It requires helping students use their knowledge and skills to understand, appreciate, and grapple with important ideas as they develop a depth of understanding for a wide range of issues and questions. Yet teaching aimed at these important goals is largely absent from U.S. classrooms."Instructional conversations" (ICs) might be one way to achieve the ambitious but elusive goals long held by many thoughtful educators. ICs are discussion-based lessons geared toward creating opportunities for students' conceptual and linguistic development. They focus on an idea or a students. The teacher encourages expression of students' own ideas, builds upon information students provide and experiences they have had, and guides students to increasingly sophisticated levels of understanding. In contrast to more directive forms of instruction, which assume that what is to be learned by the students is already in the head of the teachers, ICs assume that students themselves play an important role in constructing new knowledge and in acquiring new understandings about the world.Conversations that instruct and stimulate thinking might be particularly important for language minority students, many of whom receive insufficient opportunities for conceptual and linguistic development at school. By providing students with opportunities to engage in interactions that promote analysis, reflection, and critical thinking, instructional conversations suggest a way to help redress the imbalance of a curriculum that is heavily weighted toward skills and knowledge acquisition.

An interactive visual language for term subsumption languages

Abstract: A visual language is defined equivalent in expressive power to term subsumption languages expressed in textual form. To each knowledge representation primitive there corresponds a visual form expressing it concisely and completely. The visual language and textual languages are intertranslatable. Expressions in the language are graphs of labeled nodes and directed or undirected arcs. The nodes are labeled textually or iconically and their types are denoted by six different outlines. Computer-readable expressions in the language may be created through a structure editor that ensures that syntactic constraints are obeyed. The editor exports knowledge structures to a knowledge representation server computing subsumption and recognition, and maintaining a hybrid knowledge base of concept definitions and individual assertions. The server can respond to queries graphically displaying the results in the visual language in editable form. Knowledge structures can be entered directly in the editor or imported from knowledge acquisition tools such as those supporting repertory grid elicitation and empirical induction. Knowledge structures can be -exported to a range of knowledge-based systems.

Indexing hypertext documents in context

Abstract: To generate intelligent indexing that allows context-sensitive information retrieval, a system must be able to acquire knowledge directly through interaction with users. In this paper, we present the architecture for CID (Computer Integrated Documentation), a system that enables integration of various technical documents in a hypertext framework and includes an intelligent browsing system that incorporates indexing in context. CID's knowledge-based indexing mechanism allows case-based knowledge acquisition by experimentation. It utilizes on-line user information requirements and suggestions either to reinforce current indexing in case of success or to generate new knowledge in case of failure. This allows CID's intelligent interface system to provide helpful responses, even when no a priori user model is available. Our system in fact learns how to exploit a user model based on experience (from user feedback). We describe CID's current capabilities and provide an overview of our plans for extending the system.

An approach to knowledge acquisition based on the structure of personal construct systems

Abstract: A research effort aimed at the development and unification of the prerequisite underlying theoretical foundations for an adequate approach to knowledge elicitation from repertory grid data is described. A theory of confirmation that incorporates the basic tenets of personal construct psychology directly into the logic as a basis for the determination of relevance is offered, thus strengthening the logic and extending personal construct psychology. These largely theoretical developments are applied to the representation and analysis of repertory grid data. The concept of an alpha -plane is introduced as a binary decomposition of repertory grid data that furnishes the realization of construct extensions (or ranges of convenience) needed to determine the range of relevance of a particular generalization or hypothesis. In addition, they provide the uniquely determined string of incidences required by any application of Bundy's truth functional incidence calculus. The theories are applied to the design and construction of NICOD-a semiautomated medical knowledge acquisition system. The system has been successfully employed in the elicitation of valuable heuristic radiological knowledge (mammography) that the domain experts (radiologists) were otherwise unable to articulate. >

Effects of Learner-Control versus Program-Control Strategies on Computer-Aided Learning of Chemistry Problems: For Acquisition or Review?.

Abstract: An assumption of computer-aided learning (CAL) is that learners are the best judges of their learning needs. This assumption was examined by comparing learner-control (LC) strategies with program-control (PC) strategies. It was hypothesized that the effectiveness of the control strategies is contingent on both the learning phase (knowledge acquisition vs. knowledge review) and individual differences in prior knowledge. Fifty-five 11th graders worked through a set of 4 CAL modular tasks varying in control strategies (PC or LC) and learning phase (acquisition or review). There was a significant crossover interaction invloving the 2 experimental factors. The positive difference betwen PC and LC (in favor of PC) for knowledge acquisition was reserved for knowledge review. However, this relationship dependend on students' prior knowledge

Using a description classifier to enhance knowledge representation

Abstract: A specialized technology for reasoning with descriptions that extends the class of useful inferences beyond simple inheritance has been developed as part of the Strategic Computing Initiative of the US Dept. of Defense's Defense Advanced Research Projects Agency. Loom and the Knowledge Representation Editing and Modeling Environment (KREME), which capture this technology in the form of an inference engine, called a classifier, are described. These systems share the notion that the ability to define and reason with descriptions is basic to the task of knowledge representation. Loom supports a description language (the frame component) and a rule language, and uses its classifier to help bridge the gap between the two. The classifier gives Loom the additional deductive power to provide inference capabilities not found in current knowledge-representation tools. KREME, a knowledge acquisition and editing tool, uses a classifier to help knowledge engineers maintain consistency while developing knowledge bases. One of KREME's functions is to bring both actual and potential conflicts to developers' attention. KREME can also help merge separately developed knowledge bases. Description unification is compared with Prolog-style unification, and the benefits of description technology are examined. >

Arrangement for automated troubleshooting using selective advice and a learning knowledge base

Abstract: AN ARRANGEMENT FOR AUTOMATED TROUBLESHOOTING USING SELECTIVE ADVICE AND A LEARNING KNOWLEDGE BASE Abstract Troubleshooting expert systems are generally embodied in software for the purpose of solving difficult problems in some narrow domain of expertise. The prior art describes certain mechanics for developing or generating rules. That process is commonly known as the knowledge acquisition process. Having acquired the knowledge, our new troubleshooting arrangement eliminates the prior art separation between the expert system knowledge acquisition process and the expert system utilization process. Our new arrangement also detects and classifies invalid actions or other errors of the user in a manner that allows for the non-human expert system to advise the human user. Our arrangement interactively communicates between a user and a troubleshooting system, generates a learning knowledge base, identifies an object being tested by the user, utilizes the learning knowledge base for troubleshooting the test object, and classifies the test object as faulty or not faulty. Our arrangement structures prior actions of the user as pairs of elements in a multi-branch tree in the learning knowledge base and the compares present actions of the user with the prior actions of the user. A current signal path is generated between a first input and a first output in the system under test. Then, it can be determined whether the test object belongs to the current signal path. Responsive to the comparison of the actions of the user andresponsive to whether the test object belongs to the current path, the skill level of the user can be classified into one of a plurality of skill levels. Responsive to the skill level classification of the user, the kind of troubleshooting advice to be given the user can be classified as to one of a plurality of kinds of troubleshooting advice and thereafter the advice can be communicated to the user. Further, the output of the test object can be measured and a new current signal path can be generated and re-generated for identifying a faulty object in the system under test.