The vocabulary problem in human-system communication

Reviewer: Richard S. Marcus

.abstract In almost all computer applications, users must enter correct words for the desired objects or actions. For success without extensive training, or in first-tries for new targets, the system must recognize terms that will be chosen spontaneously. We studied spontaneous word choice for objects in five application-related domains, and found the variability suprisingly large. In every case two people favored the same term with probability < 0.20. Simulations show how this fundamental property of language limits the success of various design methodologies for vocabulary-driven interaction. For example the popular approach in which access is via one designers favorite single word will result in 80&#8211;90 percent failure rates in many common situations. An optimal strategy, unlimited aliasing, is derived and shown to be capable of several-fold improvements. &#8212; Authors Abstract This is an excellent paper for anyone interested in interface design where users choice of words is involved, especially those who believe that vocabulary is not a problem. The authors show that a few aliases (synonymous terms) can improve the success of spontaneous selection markedly, and they suggest that unlimited aliasing is the optimum solution. Three approaches to identify good alternate terms are suggested: (1) having a few users supply a &#8220;fair number&#8221; of terms apiece (say, 3&#8211;6); (2) extracting words from the text of descriptions of objects (a la full-text indexing of documents); and (3) adaptively, by noting what new terms users attempt to apply in operation of the system. The authors recognize that there is an imprecision problem, in that one term can be selected by different users to mean different objects. However, they point out that many aliases may be more precise terms than the common terms for which they substitute and thus may actually improve precision. In any case, the authors point out that there are techniques for managing the ambiguities. Their preferred method is interactively to display choices, ordered by frequency of occurence, for the user in order to enable disambiguation. The authors note that effective disambiguation may require good system explanations, which is a problem in itself. The authors also note other possible disambiguation methods (multiterm Boolean expressions, formal query languages, and natural language understanding) but turn away from these as being difficult to implement and not very successful. While I recognize the cogent analysis of much of this paper, I might question the strong emphasis on spontaneous selection. Perhaps some pre-selection mediation by the system (e.g., via menus) could avoid much post-selection disambiguation. More generally, the authors apparent aversion to considering a combination of methods in approaching this problem is questionable, although I recognize that any one of those denigrated by the authors may be inferior to unlimited aliasing as a single solution.