Eric Mays

Context based spelling correction

Abstract Some mistakes in spelling and typing produce correct words, such as typing “fig” when “fog” was intended. These errors are undetectable by traditional spelling correction techniques. In this paper we present a statistical technique capable of detecting and correcting some of these errors when they occur in sentences. Experimental results show that this technique is capable of detecting 76% of simple spelling errors and correcting 73%.

K-Rep system overview

The K-Rep system was built to explore the utility of a KL-One style knowledge representation in the development of expert systems. Beginning in about 1985, our activity in expert systems has been centered on the FAME (FinAncial Marketing Expertise) system[4]. FAME attempts to provide support to an IBM marketing representative in the financing decisions involved in the acquisition of large mainframe computer systems. Based on our experience in building a feasibility demonstration of FAME using a rule based approach, we concluded that the rule based technology would not easily scale up. Thus we abandoned the rule based approach in favor of organizing the system as a set of problem solvers around a common conceptual core. Since diverse problem solvers would be utilized in FAME, it was thought desireable that the conceptual core have a well-defined, enforceable semantics. These considerations led us to the KL-One[3] style knowledge representation.

An examination of undetected typing errors

Abstract We examine the effect of increasing word list size on the error rate of spelling correctors. An experiment on a large body of text shows that an increase in the word list size decreases the error rate.

Experience with K-REP: an object-centered knowledge representation language

An object-centered knowledge representation language is presented, with examples of its use in the FAME (financial marketing expertise) system. This representation, based on the tradition of frames and semantic nets, provides automatic reclassification, value caching, multiple knowledge bases, and an interface to relational databases. The influence that the representation has had on the development of the FAME system is discussed.<<ETX>>

Integrating rules and inheritance networks in a knowledge-based financial marketing consultation system

The integrated use of rule-based inference and an object centered knowledge representation (inheritance network) in a financial marketing consultation system is described. The rules provide a highly flexible pattern match capability and inference cycle for control. The inheritance network provides a convenient way to represent the conceptual structure of the domain. By merging the two techniques, the financial computation can be shared at the most general level, and rule inference is carried out at any appropriate level of generalization. Since domain knowledge is represented independently from control knowledge, knowledge about a particular problem-solving technique is decoupled from the conditions for its invocation. A large financial marketing system has been built, and examples are given.<<ETX>>

An experiment in constructing an open expert system using a knowledge substrate

This paper discusses an experiment in the use of an object-centered knowledge representation service to provide a common conceptual model for the construction of a large knowledge-intensive decision support tool. A core knowledge substrate forms a common resource for a variety of problem-solving activities and a basis for the rapid construction of new capabilities. FAME, a substantial expert system to aid in the financial marketing of IBM mainframes, has been built and extensively tested in the field to validate our tools and techniques.

A persistent store for large shared knowledge bases

Techniques for managing problems associated with the scalability of large knowledge-based systems are presented. The discussion is based on experience in building a large knowledge-based system and on perceptions regarding future technological requirements to support ongoing development. Achieving persistence for knowledge bases (KBs) is investigated. Persistence refers to storing a knowledge base on a stable storage medium such as a magnetic disk. A knowledge base management system (KBMS) in which a large KB is concurrently developed by a team of collaborating knowledge engineers is proposed. At the heart of the KBMS is a version store, which is a persistent storage structure for a KB. To support the concurrent collaborative work, the version store maintains multiple versions of a KB such that a knowledge engineer can access and modify any version. Retrieval and updating operations have been defined on the version store to efficiently access and modify any version. Objects in a version store are clustered to support efficient access of an entire version of the KB or subparts of it. The retrieval algorithm has been validated through simulation. A prototype of the version store has been implemented and is being integrated into the user interface.<<ETX>>