HP-RL: an expert systems language - Hewlett-Packard Representation Language - includes related article on HP-RL

Hewlett-Packard Journal, August, 1988 by Steven T. Rosenberg

HP-RL: An Expert Systems Language

FOR SEVERAL YEARS, the expert systems department of HP Laboratories investigated knowledge representation and reasoning techniques in artificial intelligence (AI). One of the ways we approached this was through the construction of an expert systems language called HP-RL (Hewlett-PAckard Representation Language). Our goal has been to create a powerful and integrated collection of tools that is usable over a wide range of application domains.

HP-RL is now a mature experiment, and work on it has come to an end. Much of the research involving applications has already been published. This paper complements previous articles describing applications in whose development HP-RL played a part and presents a retrospective look at the HP-RL research effort.

HP-RL has been supported in the past for experimental use by interested projects within HP, and has been distributed to selected participants in HP's U.S. $50,000,000 University Strategic Grants program, where it has been used in teaching graduate students and as an aid in research. Over the years, groups at HP Laboratories and HP divisions have used HP-RL to construct a variety of applications including smart software tools,.sup.1 intelligent instrumentation,.sup.2 and natural language processing..sup.3 Fig. 1 shows a partial list of the types of HP-RL-based expert systems experiments conducted at HP. We have used the feedback from these experiments to evolve the language to its final state. In addition, through the University Strategic Grants program, we received feedback from external users over a different range of uses. The result of this engineering loop has been the evolution of a robust and powerful expert systems technology that has been applied to a wide range of problem areas of interest to HP. Although HP-RL is a completed experiment, the increased availability of third-party tools and HP Prolog on the HP 9000 Series 300 Computers provides a variety of commercially available tools for those who wish to experiment with expert systems technology.

Expert systems languages such as HP-RL are part of an evolution in computer science to increasingly higher-level programming languages. Expert systems programming is characterized by the attempt to take maximum advantage of the knowledge and problem-solving skills of human experts. Consequently, expert systems programming languages are designed to simplify the task of translating an expert's knowledge into computer-usable form. The goal is to raise the level of description possible in languages such as C or Pascal through the use of various constructs for knowledge representation and reasoning.

HP-RL is an example of a modern high-end, frame-based integrated expert systems tool. Commercial products with similar functionality include ART, KEE, KnowledgeCraft, and GoldWorks. Today, integrated knowledge representation and reasoning tools are the predominant paradigm for expert systems applications. Frame-based solutions to providing this integration have been the most favored and widely used approach. HP-RL is one of many expert systems languages that have emerged in recent years as useful tools for supporting the development of expert systems applications.

HP-RL contains a frame-based component to support knowledge representation, a rule-based component to support reasoning, and a powerful query language. It supports a range of functionality and a variety of programming styles. For example, frames (the basic unit used in knowledge representation) support message passing that is consistent with object-oriented programming in Common Lisp, and the rule-based inference engine supports both backward chaining and forward chaining reasoning paradigms. In addition, a great deal of customization and modularization is possible. The frame-based knowledge representation component can be loaded and used without the rule-based inference engine. The user has the ability to declare away functionality to customize a simpler, more convenient system. On the other hand, the user can also customize a more sophisticated solution. For instance, the user can define an individual algorithm for inheritance to control the use of abstraction hierarchies in the knowledge representation component, and can also define an individual strategy for controlling reasoning.

A part of our philosophy has been that an experienced programmer needs a variety of programming tools, some of which are expert systems tools, while others are more conventional programming tools. Like a craftsman, the programmer needs the flexiblity of choosing which tools to use for a particular task. While HP-RL can function as an expert systems programming shell, we prefer the toolbox metaphor, in which various tools from the expert systems drawer can be readily intermixed with Common Lisp, object-oriented programming in Common Lisp, or conventional programming languages. Historically, most of the applications built using HP-RL have been heterogeneous in nature, so our toolbox approach has proven to be fruitful.