Browse Theses

The aim of this dissertation is to make developing knowledge-intensive applications easier by providing a methodology for designing their knowledge base. This is done by providing a very expressive, graphical knowledge base which is customized for each application. Each application has its own user-defined data model to specify access to the knowledge base. The application-specific data models allow knowledge to be shared between applications, yet access is tailored to the current task. The application developer uses our knowledge base design tool, called scWEAVE, to design new knowledge base data models and develop prototype knowledge bases which use those data models.

The application developer defines each knowledge base with type constructors and abstract data types from constructive type theory, methods on the types, and declarative graph constructors which specify how the knowledge is to be stored and retrieved from the graphical knowledge base provided by scWEAVE. scWEAVE then sets up an interface to the provided graphical knowledge base which appears as a prototype knowledge base with the specified data model. The knowledge base is accessed through an extensible knowledge base programming language that allows general views to be created for definition, manipulation, and retrieval of the knowledge.

To guide development of the application-specific knowledge base and its data model, we have developed a process for knowledge base design which is incorporated in scWEAVE. We take advantage of sophisticated theoretical tools which have been proven effective in other areas of computer science and extend them to form a strong, theoretical foundation for knowledge base design. This allows scWEAVE to create prototype knowledge bases from a high-level specification given by the application developer. We import formalisms from knowledge representation, natural language semantics, programming language research, constructive type theory, and databases.

We give the theoretical foundations for knowledge base design and explain how we have used them to cleanly implement the tool scWEAVE which incorporates them. We demonstrate our approach by developing knowledge base data models which are useful in the areas of general knowledge representation, problem solving, natural language processing, object-oriented databases, and molecular biology. We show how scWEAVE supports our process for knowledge base design by developing example applications with emphasis on a realistic application from human genetics, specifically integrating order and distance information from genetic linkage, physical, and radiation hybrid genome maps.