The main contributions of this thesis are a Bayesian theory of learning classification rules, the unification and comparison of this theory with some previous theories of learning, and two extensive applications of the theory to the problems of learning class probability trees and bounding error when learning logical rules. The thesis is motivated by considering some current research issues in machine learning such as bias, overfitting and search, and considering the requirements placed on a learning system when it is used for knowledge acquisition. Basic Bayesian decision theory relevant to the problem of learning classification rules is reviewed, then a Bayesian framework for such learning is presented. The framework has three components: the hypothesis space, the learning protocol, and criteria for successful learning. Several learning protocols are analysed in detail: queries, logical, noisy, uncertain and positive-only examples. The analysis is done by interpreting a protocol as a...

Much of the research in inductive learning concentrates on problems with relatively small amounts of data. With the coming age of ubiquitous network computing, it is likely that orders of magnitude more data in databases will be available for various learning problems of real world importance. Some learning algorithms assume that the entire data set fits into main memory, which is not feasible for massive amounts of data, especially for applications in data mining. One approach to handling a large data set is to partition the data set into subsets, run the learning algorithm on each of the subsets, and combine the results. Moreover, data can be inherently distributed across multiple sites on the network and merging all the data in one location can be expensive or prohibitive. In this thesis we propose, investigate, and evaluate a meta-learning approach to integrating the results of mul...

The knowledge acquisition bottleneck impeding the development of expert systems is being alleviated by the development of computer-based knowledge acquisition tools. These work directly with experts to elicit knowledge, and structure it appropriately to operate as a decision support tool within an expert system. However, the elicitation of expert knowledge and its effective transfer to a useful knowledge-based system is complex and involves a diversity of activities. This paper illustrates the complete development of a decision support system using knowledge acquisition tools. The example is simple enough to be completely analyzed but exhibits enough real-world characteristics to give significant insights into the processes and problems of knowledge engineering. 1 Introduction Knowledge acquisition for expert system development has come to be termed knowledge engineering, following Feigenbaum's (1980) use of the term to describe the reduction of a large body of knowledge to a precise...

The success of large software engineering projects depends critically on the specification, which must represent the requirements of a large number of people with widely differing perspectives. Conventional approaches to software engineering do not address the process of identifying and integrating these perspectives, but instead concentrate on the maintenance of a single consistent description. This results in a specification which represents only one point of view, often the analyst's, excluding suggestions which do not fit with this view. The processes which led to the adoption of this point of view will go unrecorded, making any rationale attached to such a specification incomplete. Other participants will not be able to validate it properly, as it does not relate to their requirements. This thesis integrates ideas drawn from the study of knowledge acquisition, computer-supported co-operative work and negotiation into a model of the specification activity which allows the capture ...

As knowledge-based systems are addressing increasingly complex domains, their roles are shifting from classical expert systems to interactive assistants. To develop and maintain such systems, an integration of thorough knowledge acquisition procedures and sustained learning from experience is called for. A knowledge level modeling perspective has shown to be useful for analyzing the various types of knowledge related to a particular domain and set of tasks, and for constructing the models of knowledge contents needed in an intelligent system. To be able to meet the requirements of future systems with respect to robust competence and adaptive learning behavior, particularly in open and weak theory domains, a stronger emphasis should be put on the combined utilization of casespecific and general domain knowledge. In this chapter we present a framework for integrating KA and ML methods within a total knowledge modeling cycle, favoring an iterative rather than a top down approac...

Clustering is used in information retrieval systems to enhance the efficiency and effectiveness of the retrieval process. Clustering is achieved by partitioning the documents in a collection into classes such that documents that are associated with each other are assigned to the same cluster. This association is generally determined by examining the index term representation of documents or by capturing user feedback on queries to the system. In cluster-oriented systems, the retrieval process can be enhanced by employing characterization of clusters. In this paper, we present the techniques to develop clusters and cluster characterizations by employing user viewpoint. The user viewpoint is elicited through a structured interview based on a knowledge acquisition technique, namely personal construct theory. It is demonstrated that the application of personal construct theory results in a cluster representation that can be used during query as well as to assign new documents to the approp...

Much of the research in inductive learning concentrates on problems with relatively small amounts of training data. With the steady progress of the Human Genome Project, it is likely that orders of magnitude more data in sequence databases will be available in the near future for various learning problems of biological importance. Thus, techniques that provide the means of scaling machine learning algorithms requires considerable attention. Meta-learning is proposed as a general technique to integrate a number of distinct learning processes that aims to provide a means of scaling to large problems. This paper details several meta-learning strategies for integrating independently learned classifiers on subsets of training data by the same learner in a parallel and distributed computing environment. Our strategies are particularly suited for massive amounts of data that main-memory-based learning algorithms cannot handle efficiently. The strategies are also independent of the particular...

In this paper, we advance a technique to develop a user profile for information retrieval through knowledge acquisition techniques. The profile bridges the discrepancy between user-expressed keywords and system-recognizable index terms. The approach presented in this paper is based on the application of personal construct theory to determine a user's vocabulary and his/her view of different documents in a training set. The elicited knowledge is used to develop a model for each phrase/concept given by the user by employing machine learning techniques. Our model correlates the concepts in a user's vocabulary to the index terms present in the documents in the training set. Computation of dependence between the user phrases also contributes in the development of the user profile and in creating a classification of documents. The resulting system is capable of automatically identifying the user concepts and query translation to index terms computed by the conventional indexing process. The ...

this document for granted, concentrating instead on the downstream areas of software development. In this chapter, we argue that the problems of requirements engineering deserve greater study. To understand why this is so, we consider the role of the specification in the software engineering process, and describe issues which must be addressed during specification construction. The difficulties of requirements engineering come from many directions, including the sheer quantity of knowledge involved, the inherent uncertainty, and the need for negotiation where there are conflicting requirements. We conclude that a prescriptive framework to support negotiation of requirements is highly desirable, and describe a number of objectives for such a framework.

Despite the progress made in the field of intelligent tutoring systems (ITS), it is still a major challenge to build systems that can teach about complex, ill-structured domains. A chief reason is that detailed, dynamic modeling of students ' knowledge is intractable in such areas, and complete, correct models of expert knowledge are inherently difficult to build. These difficulties have led some to argue that the goal of intelligent tutoring should be abandoned and that more benefit could be provided by systems without tutoring. We believe that there are many areas in which tutorial intervention is essential, particularly for the communication of expertise. In this paper we advocate basing tutorial intervention on an analysis of a student's likely points of interest within a learning environment, rather than on his or her state of knowledge. This interestfocused approach results in considerable simplification of the modeling task, and has other advantages as well. We describe an interest-tracing intelligent tutoring framework that we have been using to build learning environments for such ill-structured tasks as selling, managing, and other interpersonal skills using tutorial guidance. Our design is based on case-based reasoning as a model of human problem-solving. Expertise is modeled as an organized library of cases; student modeling is restricted to the considerations that enter into the decision to retrieve and present relevant cases. This paper describes the cognitive theory underlying our tutoring approach, and the implementation of the tutor. We show how it is possible to present useful tutorial intervention based on a student's state of interest, without an overwhelming burden of student modeling.