Word prediction strategies in program editing environments

Abstract

The goal of this research is to demonstrate that word prediction in a constrained domain like a programming language can be done effectively and hence has the potential to facilitate typing for disabled people who are motor-impaired with regard to typing. This paper presents a set of word prediction strategies for a programming editor. In order to evaluate the effectiveness of these strategies, a system which simulates a Turbo Pascal editor was developed. The strategies are found to be efficient and improve the accuracy of word prediction to various degrees. Although the system is a simulator of a Turbo Pascal editor, the applicability of each of the strategies is not confined to the syntax of this language and, therefore, they can be implemented on any language-directed editors. The results could also have relevance for other text generation systems where the aim is to save the user effort.