http://hdl.handle.net/2142/1209 27th Clinic on Library Applications of Data Processing. Edited by F.W. Lancaster and Linda C. Smith http://hdl.handle.net/2142/1303 Walking Your Talk: Why Information Managers are Not High Tech Penniman, W. David This paper discusses the role of technology in creating successful information services, and also the important role of people in creating successful implementations of new technologies. Much has been written about the areas of artificial intelligence and expert systems. This paper will try to stay on a broader level, its ideas applying to a wide range of technologies, well beyond the traditional library arena. Although most readers of these proceedings are concerned with information management within the library context, libraries provide an environment well suited for a more general discussion about change and the role of technology in introducing change necessary for survival. Libraries --Automation Expert systems (Computer science) Artificial intelligence Library science --Data processing http://hdl.handle.net/2142/1302 Expert Systems in Document Delivery: The Feasibility of Learning Capabilities Pontigo, Jaime Tovar-Reyes, Ezequiel Rodriguez, Guillermo Ortiz-Gama, Sergio To solve the problem of document delivery in Mexico, the authors developed SEADO (Expert System for Document Supply). SEADO consists of three main components: a knowledge base, an expert system shell, and the database. The knowledge base was built through fault tree analysis and through structured flowcharts. The shell was developed with EXSYS, a generalized expert system development package. The database was based on information sources of various kinds: printed material, local databases, public databases, etc. To evaluate the impact of different learning capabilities, the authors decided to test alternative ways of achieving a predictor for the system to perform in a dynamic and adaptive way. Learning by a weighted-based scheme was compared with a probability-based scheme. Libraries --Automation Expert systems (Computer science) Artificial intelligence Library science --Data processing http://hdl.handle.net/2142/1301 Intelligent Interfaces to Online Databases Vickery, Brian C. The possible functions of intelligent interfaces are summarized. Five examples of recent or current European projects on the development of interfaces are described: INSERM, GIRT, EURISKO, ERLI/ MINITEL, and IMIS. A number of the problems of interface design and implementation are reviewed. Libraries --Automation Expert systems (Computer science) Artificial intelligence Library science --Data processing http://hdl.handle.net/2142/1300 Knowledge Representation in Artificial Intelligence Travis, Irene L. The problems of knowledge representation and use in expert systems and the problems of organizing and searching information in libraries and other bibliographic systems have much in common. There are two basic paradigms for representing knowledge in the knowledge bases of expert systems: rule-based and object-based. Of the two, the rulebased approach has had more publicity, but the object-oriented approach, which will seem more familiar to librarians, is coming to be seen as a necessary complement to rules or even as the more basic system component. One of the principal unsolved problems in knowledge representation is how to provide expert systems and natural language processing systems with more world knowledge, particularly "common sense" knowledge, in order to make them more robust. A major project to build a knowledge base of such basic information is underway at the Microelectronics and Computer Corporation (MCC), a corporation financed by a consortium of American industry to carry out research in advanced computing and computing technology. Since the project represents an attempt to organize a very large and general body of knowledge for use, it can be hypothesized that it will face many of the same problems faced by librarians as they have done the same thing. The project's published goals and achievements at the midpoint of its ten-year life are reviewed from that perspective. Four barriers to such efforts are discussed: (1) the variability of human performance in tasks related to knowledge representation and search; (2) the paradox of structure; (3) the double-edged nature of the 80/20 rule; and (4) the inertia of an installed base. Libraries --Automation Expert systems (Computer science) Artificial intelligence Library science --Data processing http://hdl.handle.net/2142/1299 Natural Language Processing: Current Status for Libraries Warner, Amy J. A general introduction to natural language processing is provided, including a definition and an overview of how natural language processing systems work. Representative systems from both the research and applied sectors are presented in order to illustrate the state of the art in the field and the issues which underlie system design and implementation. Actual and potential areas for natural language processing in information retrieval, including retrieval from online catalogs, indexes, and full texts are discussed, with an assessment of short- and long-range agendas and possible limitations. Libraries --Automation Expert systems (Computer science) Artificial intelligence Library science --Data processing http://hdl.handle.net/2142/1298 User Models for Information Systems: Prospects and Problems Borgman, Christine L. Plute, Yolanda I. Expert systems attempt to model multiple aspects of human-computer interaction, including the reasoning of the human expert, the knowledge base, and characteristics and goals of the user. This paper focuses on models of the human user that are held by the system and utilized in interaction, with particular attention to information retrieval applications. User models may be classified along several dimensions, including static vs. dynamic, stated vs. inferred, and short-term vs. longterm models. The choice of the type of model will depend on a number of factors, including frequency of use, the relationship between the user and the system, the scope of the system, and the diversity of the user population. User models are most effective for well-defined tasks, domains, and user characteristics and goals. These user-system aspects tend not to be well defined in most information retrieval applications. Libraries --Automation Expert systems (Computer science) Artificial intelligence Library science --Data processing http://hdl.handle.net/2142/1297 Expert Systems at the National Agricultural Library: Past, Present, and Future Waters, Samuel T. Since 1986, the National Agricultural Library (NAL) has developed four expert advisory systems for ready reference on agricultural topics, and has trained librarians from other institutions who have contributed three other systems. All may be downloaded from the NAL electronic bulletin board. NAL has stimulated development elsewhere of several reference advisors in subjects other than agriculture, and has actively promoted interest in the use of expert systems in libraries. NAL has been responsible for the development of an "intelligent document" in the field of aquaculture, which uses hypertext and contains an expert system. Similar products in several other fields are underway. "Smart" courseware is also being developed for library training. In the future, NAL may explore the use of new artificial intelligence techniques such as neural networks, will increase development of multimedia products and use of multifunctional software, working toward the development of knowledge access and utilization systems in important areas of concern in agriculture. Libraries --Automation Expert systems (Computer science) Artificial intelligence Library science --Data processing http://hdl.handle.net/2142/1296 Reference Expert Systems: Foundations in Reference Theory Parrott, James R. A reference expert system may be considered to be a system with a knowledge base covering various aspects of the reference process in a library setting. Knowledge bases generally consist of several components (such as databases, rule bases, frames, and semantic nets) that interact with an inference engine, a user interface, and each other. This paper will examine progressively more complex knowledge-based systems for reference that can be constructed from components like these, concentrating at first on combinations of databases and rule bases. This examination will lead to a classification of reference expert systems. In Section 3, very simple architectures of a type common in other fields will first be considered. Arguments drawn from reference theory suggest that these simple architectures are appropriate primarily in dealing with directional reference transactions. In Sections 4 and 5, reference theory will be used to develop two additional architectures more appropriate to other reference transactions, such as ready-reference transactions. The classification of reference expert systems will be completed in Section 6 by examining further reference theory and then using it to develop variants on the three basic types. Section 7.1 will discuss briefly the use of reference knowledge bases for computer-assisted instruction. Section 7.2 will consider deep reference knowledge. The paper will conclude with some prognostications about future developments. Libraries --Automation Expert systems (Computer science) Artificial intelligence Library science --Data processing http://hdl.handle.net/2142/1295 Interactive Knowledge-Based Systems for Improved Subject Analysis and Retrieval Humphrey, Susanne M. A knowledge-based, or expert, system encoding both factual and procedural knowledge to assist users in performing an intellectual task is ideally suited to indexing. Existing thesauri, classification schemes, and indexing manuals are a good starting point, and artificial intelligence (AI) computer languages and data structures seem well suited for development of these systems. In addition, currently available workstation environments (with windows and mouse) and standard software (such as X Windows) should make possible sophisticated and portable interfaces. A unique prototype, the MedlndEx System, is being developed to assist people using the Medical Subject Headings (MeSH ) thesaurus to index the MEDLINE database in MEDLARS (Medical Literature Analysis and Retrieval System) at the National Library of Medicine (NLM). MedlndEx is, in principle, applicable to any indexing system using a thesaurus and following a body of indexing rules. Libraries --Automation Expert systems (Computer science) Artificial intelligence Library science --Data processing http://hdl.handle.net/2142/1294 Automated Cataloging: Implications for Libraries and Patrons Weibel, Stuart L. Many changes in cataloging practice have been and will increasingly be technology driven. Bound lists and drawers of cards defined the form and function of catalogs for as long as they existed until the advent of digital computers. Even today, however, MARC records are as much a derivative of catalog cards as the reverse. The additional functionality of computer catalogs affords opportunities to increase the effectiveness of the cataloging process and improve the value of the catalog itself. Three main research areas are examined with regard to their anticipated influence on this evolution. Automated cataloging research, focusing on the application of rule-based systems to cataloging, represents a novel way to address the cataloging process per se, but has as yet made only modest progress. The incremental implementation of a variety of computer-assisted methods for addressing aspects of cataloging represents a second, more conventional approach to advancing the state of the art in cataloging automation. This approach shares the goal of the first to build intelligent capabilities into cataloging systems but the focus remains on human cataloging systems and the methods of implementation are more conventional. The third area is not part of traditional concepts of cataloging at all, but will have a major impact upon what is available in catalogs in the broadest sense of that term. This "non traditional" cataloging involves automated processing of documents to extract bibliographic information as well as full text. It will expand the range of cataloged objects to include items not generally cataloged due to resource constraints. Automated processing of such materials will be characterized by lower quality and less complete cataloging, but will nonetheless promote improved access to materials that are currently lost to bibliographic control. Libraries --Automation Expert systems (Computer science) Artificial intelligence Library science --Data processing