Computers for Composition: A Stage Model Approach to Helping

Abstract

This paper is concerned with how computer can assist text composition. It is a review of what has been or is being done — particularly in the EPISTLE project at the Yorktown IBM Watson Research Center — and it is also a preview of what could be done in the future. The discussion is centered around a stage model of composition adapted from computer programming. The model’s psychological validity, although credible, is secondary to the rather natural way it appears to organize possible computer functions supporting composition. Some stages of this model clearly require powerful language technologies, and the stronger interest is clearly in these. Nevertheless, a number of functions are identified for other stages, which functions do not require extensive linguistic capability to implement or enhance. Throughout the paper there is a deliberate bias towards "practical" writing — writing whose fruits are of unquestioned value and whose authors receive monies undetermined by word counts or reviewers’ praise. The notion of this type of writing — most frequent by far — is that the composition meets a requirement of business or social commerce, whether it is an interagency status report, product documentation, a step in the process of attempting to free a client, or a worded advertising fantasy designed to attract cash customers. The paper concludes with some consideration of specialized audiences for the various computer functions as well as the roles educators might play in promoting (or obstructing) their development.