An evaluation of editing paradigms

Abstract

The choice between tree-building and text-recognition paradigms has been an issue in language-based editor design over the past decade, with much intuitive comment appearing in the literature. To the best of our knowledge, however, no systematic attempt to demonstrate the advantage of either paradigm, by application of relevant theories or by controlled experimental evaluation, has been attempted. The problem is complicated by the interaction of several factors -- in addition to the user's conceptual model of a structured document, factors such as the error discipline to be applied, the model bias produced by textual display, and the adequacy of implementation of the text-recognition approach, have all to be taken into account. In this theoretical study we analyse and compare various editing tasks undertaken by software engineers. The Keystroke Level Model (KLM) is used to assess the efficiency of the paradigms.