Issue No. 03 - June (1996 vol. 11)
DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/64.506749
<p>Wide-area networks and the internet-based World Wide Web allow developers to provide intelligent knowledge servers. Expert systems running on servers can support a large group of users who communicate with the system over the network. In this approach, user interfaces based on Web protocols provide access to the knowledge servers. Users do not need special hardware or software to consult these services with appropriate Web browsers.</p> <p>Providing expert advice to a large group of end users has proved surprisingly difficult. In the eighties, many researchers and practitioners predicted that expert-system technology would lead to the development of many small and medium-sized advisory systems that could help many categories of novice users in performing expert-level tasks. Although technical progress in the last decade has been satisfactory, the number of widespread expert systems in use is disappointing. Because they circumvent the usability problem, embedded expert systems, however, tend to be more successful than advisory systems. The design of truly usable advisory systems has proved more difficult than initially expected.</p> <p>A major problem in making expertise readily available is that, compared to other applications such as word processors, users typically run advisory systems infrequently. This usage pattern stems from the relatively few critical cases users cannot handle normally. Therefore, developers must design advisory systems for infrequent operation. In this context, program distribution is a serious problem for many types of expert systems. Because most knowledge bases must be updated regularly, users must upgrade their software frequently. When expert systems require access to large databases, the volumes of data to be updated can be significant. The effort required to keep system installation current can overwhelm developers and users.</p> <p>The widespread use of the Internet and World Wide Web provides an opportunity for making expert systems widely available. By implementing expert systems as knowledge servers that perform their tasks remotely, developers can publish expertise on the Web. Technologies and infrastructures that make this approach feasible are emerging. Simultaneously, interest in artificial intelligence support for network navigation services is growing. Several research groups are working on intelligent agents that help users find information and perform other tasks on large networks. We can view these agents as expert systems that incorporate knowledge on how to find appropriate network services.</p> <p>To make knowledge servers available, developers must distribute the software front ends that allow users to communicate with the servers. The Web supports the Common Gateway Interface, which can provide form-based front ends to databases, and to expert systems with simple user interfaces. Currently, developers are experimenting with CGI-based programs for various services. This approach, however, does not allow user interaction beyond form filling and the use of Web documents as program output.</p> <p>Sun Microsystems' Java programming language, conversely, provides a powerful basis for implementing user interfaces to knowledge servers: Developers can include programs written in Java in HyperText Markup Language (HTML) documents; Web browsers can download these programs over the Internet and run them locally. Java programs can then act as user interfaces to expert systems by opening network connections to knowledge servers. This article will show how developers can design and implement such systems, and will describe a prototype implementation of an expert system with a Java-based user interface.</p>
H. Eriksson, "Expert Systems as Knowledge Servers," in IEEE Intelligent Systems, vol. 11, no. , pp. 14-19, 1996.