This special issue of the International Journal of Intelligent Systems is concerned with the use of knowledge-based technique? for information retrieval (IR). As an area of research, IR has a long history in which the main concerns have been how to represent the contents of documents so that a user's request can be matched to those documents, how to provide "simple to use" query languages which nevertheless have sufficient expressive power, and, importantly, how to develop and conduct performance evaluations of the tools and techniques that have emerged from the research effort. A number of well known techniques (e.g., Boolean, vector, and probabilistic methods) have been developed, explored, and refined, so that today there are several commercially available tools for document retrieval based on these ideas. * Nevertheless, with the increasing use of personal computers, advanced communications capabilities, and low-cost optical storage, there has been a growing recognition that these well-established techniques do not always satisfy the needs of prospective users of the information contained in the documents. A possible solution is to make use of knowledge-based techniques developed as part of the last decade's search for artificial intelligence. Such techniques might be used in a number of ways in IR, but in this issue the focus is on the kind of knowledge needed, and the way in which it is to be represented.
In assembling this issue, I have selected five articles that describe recent interesting work, and hope that the readers of the International Journal of Intelligent Systems will be stimulated by this challenging application area. Of course, this selection is not an exhaustive review of current research in knowledge-based information retrieval. For example, the work by Hafner3 on legal information retrieval, the work by Humphrey4 on medical information retrieval, and thc work by Krawczak et al.' on a system for bibliographic search of the environmental pollution literature, represent interesting, knowledge-based solutions to specific IR problems. In the first article, Weaver et al. describe their attempts to define a single, *For those who would like to read more deeply in the area of general IR, standard references are Salton and McCill' and van Rijsbergen.'