Large Knowledge-based Systems Research Articles

Modeling the Evolution of Knowledge in Learning Systems

How do reasoning systems that learn evolve over time? What are the properties of different learning strategies? Characterizing the evolution of these systems is important for understanding their limitations and gaining insights into the interplay between learning and reasoning. We describe an inverse ablation model for studying how large knowledge-based systems evolve: Create a small knowledge base by ablating a large KB, and simulate learning by incrementally re-adding facts, using different strategies to simulate types of learners. For each iteration, reasoning properties (including number of questions answered and run time) are collected, to explore how learning strategies and reasoning interact. We describe several experiments with the inverse ablation model, examining how two different learning strategies perform. Our results suggest that different concepts show different rates of growth, and that the density and distribution of facts that can be learned are important parameters for modulating the rate of learning.

Graph Traversal Methods for Reasoning in Large Knowledge-Based Systems

Commonsense reasoning at scale is a core problem for cognitive systems. In this paper, we discuss two ways in which heuristic graph traversal methods can be used to generate plausible inference chains. First, we discuss how Cyc’s predicate-type hierarchy can be used to get reasonable answers to queries. Second, we explain how connection graph-based techniques can be used to identify script-like structures. Finally, we demonstrate through experiments that these methods lead to significant improvement in accuracy for both Q/A and script construction.

When innovating, go slow [Spectral Lines

THE TANGLED HISTORY of innovation reveals a peculiar lesson: Slow is often better than fast. The current assumption is that innovation at its best hits like a hurricane. Austrian-American economist and political scientist Joseph Schumpeter, who first recognized the importance of innovation for economic growth, famously described innovation as coming in gales, sweeping aside all that came before. But whether it's biomedical, digital, or electromechanical, systems-level innovation requires human ingenuity, even wisdom. And the wise adaptation of advances in technoscience—in the design, engineering, and management of large knowledge-based systems that deliver energy, information, transportation, security, food, and health—takes time.

Knowledge Base Decomposition to Facilitate Verification

We examine the verification of large knowledge-based systems. When knowledge bases are large, the verification process poses several problems that are usually not significant for small systems. We focus on decompositions that allow verification of such systems to be performed in a modular fashion. We identify a graphical framework, that we call an ordered polytree, for decomposing systems in a manner that enables modular verification. We also determine the nature of information that needs to be available for performing local checks to ensure accurate detection of anomalies. We illustrate the modular verification process using examples, and provide a formal proof of its accuracy. Next, we discuss a meta-verification procedure that enables us to check if decompositions under consideration do indeed satisfy the requirements for an ordered polytree structure. Finally, we show how the modular verification algorithm leads to considerable improvements in the computational effort required for verification as compared to the traditional approach.

CHEMREG: Using Case-Based Reasoning to Support Health and Safety Compliance in the Chemical Industry

CHEMREG is a large knowledge-based system used by Air Products and Chemicals, Inc., to support compliance with regulatory requirements for communicating health and safety information in the shipping and handling of chemical products. This article concentrates on one of the knowledge bases in this system: the case-based reasoner. The case-based reasoner addresses the issue of how proper communication of public health and safety information can be ensured while rapid and cost-effective product evaluation is allowed in the absence of actual hazard testing of the product. CHEMREG generates estimates of hazard data for new products from similar products using an existing relational database as a case library. Implementation of the case-based reasoner in rules and objects using a commercial knowledge-based system shell is described. Although some refinements remain, the performance of the case-based reasoner has met its design goals.

Software maintenance support for knowledge-based systems

With the increasing use of knowledge-based systems in public and private organizations, maintenance of these systems is becoming crucial. However, the main knowledge representation paradigms have not been carefully examined in terms of maintainability. In this article, we provide a synthesis of software engineering literature on software maintenance, review the state of knowledge-based systems development in practice, and, using a case study along with a number of real-world examples, explain why many practitioners are encountering problems in maintaining large knowledge-based systems. The results of the case study demonstrate that broader spectrum, mixed-representation languages are more in line with the software engineering maintenance principles than are single-paradigm languages.

Performance evaluation of a large diagnostic expert system using a heuristic test case generator

Validating the performance of a knowledge-based system is a critical step in its commercialization process. Without exception, buyers of systems intended for serious purposes require a certain level of guarantees about system performance. This is particularly true for diagnostic systems. Yet, many problems exist in the validation process, especially as it applies to large knowledge-based systems. One of the biggest challenges facing the developer when validating the system's performance is knowing how much testing is sufficient to show that the system is valid. Exhaustive testing of the system is almost always impractical due to the many possible test cases that can be generated, many of which are not useful. It would thus be highly desirable to have a means of defining a representative set of test cases that, if executed correctly by the system, would provide a high confidence in the system's validity. This paper describes the experiences of the development team in validating the performance of a large commercial diagnostic knowledge-based system. The description covers the procedure employed to carry out this task, as well as the heuristic technique used for generating the representative set of test cases.

Developing integrated object environments for building large knowledge-based systems

This paper discusses the development of environments for supporting large persistent knowledge-based systems, based around the concept of object common to artificial intelligence, programming languages, and object-oriented databases. A prototype, MOSS, was constructed to test the possible difficulties that one could encounter. The MOSS example is used to report how various features taken from the different fields interact when assembled in a single system, as a consequence of global design choices.

Linked-learning for knowledge acquisition: a pilot's associate case study

We developed a knowledge acquisition system that uses an Explanation-Based Learning domain theory as a knowledge repository from which general knowledge structures can be compiled and then translated by smart translators into the various specialized representations required for the separate expert system modules of a distributed pilot aiding system. We call this two-stage learning-plus-translation process linked learning. This architecture addresses learning for multiple modules with different knowledge representations and performance goals, but which must all perform together in an integrated fashion. It also addresses learning for an intelligent agent which must perform in a real-world, dynamically-changing environment with multiple sources of uncertainty. Finally, it serves as a case study offering insights into the integration of machine learning into the system engineering process for a large knowledge-based system development effort.

THE HIDDEN TREASURE

Abstract Starting from general observations on form and substance, I take a closer look at knowledge representation and how this field has been treated in artificial intelligence (Al) research. I argue for directing more attention to the substance of knowledge and find similar views expressed in Lenat and Guha's Cyc project (Building Large Knowledge-Based Systems. Reading, MA: Addison-Wesley, 1989). Contrary to their view, however, I argue for not dismissing prior work like research concerning thesauri, and stress the importance of an interdisciplinary and intercultural approach in general. I also note some caveats and observations concerning the distinction between representing the world and its representation by means of language.

ESR—A large knowledge-based system project of european power generation industry

The paper presents an integrated knowledge-based system aimed at providing technical support to engineers performing assessment and management of remaining life of the high-temperature pressurized components, mainly piping, in power plants. The system (ESR) is being developed at MPA Stuttgart (Germany) with support and under sponsorship of seven European electric utility companies. The initial requirements imposed by the utilities have been met by integrating various sources of knowledge, information, and data, as well as various information technology tools and techniques. The expert system modules use both production rules and object-oriented programming (decision/analysis support) and rely strongly on other integrated software: hypermedia (documentation base and explanation facility), numerics (engineering calculi), and data bases (material, standard geometries, etc.).

The design of large knowledge-based systems: the example of Digital Equipment's XSEL project

Abstract. This paper discusses the management problems associated with building and implementing large systems. The example described is XSEL, a configuring expert system designed by the Digital Equipment Corporation for worldwide application. Digital, like many other larger computer manufacturers, had experienced problems in achieving a high level of configuring accuracy when assembling its computer. These problems showed up in the manufacturing plants but originated in the sales offices. They caused difficulties with customers and increased manufacturing and selling costs. The company decided that an expert system could solve the problem. The processes associated with the design of this successful system created as many human as technical challenges. The paper describes these, discusses how and why they originated, and evaluates Digital's strategies for solving them. It makes some general recommendations for the successful management of major change.

Matias: Development and maintenance of a large but well-defined application

Abstract This paper presents experiences from development, use, and maintenance of a large rule-based expert system known as Matias. The application contains knowledge about farming loans in Finland. Using this large application as a reference point, we will discuss project management, knowledge acquisition, knowledge base design, and maintenance of large knowledge-based systems (KBS). Based on the experiences gained during the development of Matias we would like to promote better planning in KBS projects, use of structured knowledge base design principles, and use of graphics in representing knowledge to the experts. One graphical representation technique, dependency graphs, is presented in this paper. Techniques presented in this paper are applicable to most application areas but are of most importance to developers of large applications. Building a large application sets high requirements for system design, knowledge transparency, and knowledge management facilities.

Logic engineering for knowledge engineering: design and implementation of the Oxford System of Medicine

The Oxford System of Medicine project aims to develop a comprehensive information management and decision support system (DSS) for general practitioners. The nature of general medical practice means that such a system must eventually embody knowledge of a large proportion of contemporary medicine, support a range of information management tasks (document retrieval and search, database retrieval and patient record management) and decision making activities (including diagnosis, treatment, investigation, risk assessment and referral), and provide these services within a versatile and extensible package. The present implementation demonstrates a number of the required capabilities. The project offers an opportunity to investigate practical and theoretical issues of large scale knowledge-based systems. To meet requirements for scale, flexibility and soundness we believe that a logic engineering approach has many advantages. The paper describes the logic engineering techniques which are being used, provides a progress report on the design of the OSM, and outlines some current research and expected future developments.

Review of Building Large Knowledge-Based Systems

Building Large Knowledge-Based Systems (Addison-Wesley, Reading, Massachusetts, 1990, 372 pages, $39.75, ISBN 0-201-51752-3) by Douglas B. Lenat and R. V. Guha is an interim report on the Microelectronic and Computer Technology Corpporation (MCC) Cyc project.

Engineering large knowledge-based systems

Designers of knowledge-based systems seem to have the perception that realistic systems can be built by individual programmers. This approach no longer works for very large systems. Building large systems is a collaborative effort of people who each take responsibility for only part of the work. Software engineering has focused on the complications arising from system size and organizing people. This paper describes how software engineering might help in recognizing the problems of building large knowledge-based systems and in offering some methods tools and techniques to solve these problems. The paper ends with making the point that benefits may also flow in the other direction, from Artificial Intelligence to software engineering, resulting in a general improvement of these tools, methods and techniques.

Partitioning and composing knowledge

This paper argues for an approach which places the management of large knowledge bases into a comprehensive, engineering-oriented framework, and reports on an initial demonstration of these concepts. The underlying concepts are well-recognized as being effective in many areas of science: 1. 1. Partitioning of the knowledge into manageable segments. 2. 2. Rules for the composition of these segments. 3. 3. A language to provide access to these segments, control their composition, and provide the power of the system in a flexible and clear way. The motivation for this research is to deal with problems that are beginning to occur in large knowledge-based systems. As current developments of such systems lead to further growth, we foresee that their management needs will exceed the capabilities of the existing system infrastructure. In particular, we find that in the past issues related to knowledge maintenance have been ignored. Maintenance of knowledge-bases is critical if the systems are to persist.

Evaluating and validating very large knowledge-based systems

Most knowledge-based systems for use in medicine have been developed in response to specific problems such as the diagnosis of abdominal or chest pain in an accident and emergency department, or the diagnosis and treatment of meningitis. There is a role for a general decision support system capable of answering queries about any aspect of medicine, particularly in primary care. However, evaluating such a knowledge base requires more elaborate methodology than a simple iterative test and refine cycle. At the design stage an adequate knowledge base structure is required to allow focused modification of the knowledge base when errors are discovered. During the prolonged evaluation cycle the partially formed knowledge base must be tested with such techniques as validation checks for consistency and completeness and examination of characteristics of problem-solving procedures. Finally a variety of criteria that represent the performance, robustness, flexibility, predictability, validity, coverage, relevance and congruity of the knowledge base are needed for a full description of the system's worth. Two case studies from the Oxford System of Medicine project are provided as examples of this philosophy: validating specific medical facts and comparing two methods for aggregating reasoning for and against a decision option.

電圧/無効電力調整計画のための知識ベースシステムの開発

Voltage/reactive power planning is a tremendous process which requires a planner to have intellectual expertise. This paper presents a large knowledge-based system that can make voltage/reactive power planning. The knowledge acquisition has been completed on the basis of interview and protocol analysis methods. The result shows that the voltage/reactive power planning is composed of three phases: the first is reactive power adjustment where overall reactive power balance is investigated; the second is voltage adjustment where all bus voltages are tuned within the permissible operational range; and the third is voltage adjustment with a single fault of all major lines. The knowledge base is coded by about 700 production rules. It can succeed in making a voltage/reactive power plan for a very large power system.

Comments on “a comparison of concatenated and superimposed code word surrogate files for very large data/knowledge bases”

Direct searching and retrieval from a very large data/knowledge base systemare very time-consuming. The use of surrogate files, or description files, based on indexing techniques can speed up the process of information retrieval. In a recent paper [1], two indexing techniques, Concatenated and Superimposed Code Words, CCW and SCW respectively, were suggested for handling very large knowledge base systems with file size approaching Gigabytes. The space requirements and the query response times of surrogate files based on the two indexing techniques were estimated and compared. From the comparison, the CCW indexing technique was claimed to be better: it required relatively less storage and provided a faster response time. In this paper, the analysis leading to the claim is re-examined. Furthermore, it is shown that the comparison was unfairly conducted and it was biased against SCW. Contrary to the claim, SCW surrogate files do not necessarily require more storage than CCW surrogate files. In the first section, the analysis in [1] is briefly summarised. A revised analysis is described in Section 2. Finally, conclusions are given in Section 3.