Human Factors and Behavioral Science: Statistical Semantics: Analysis of the Potential Performance of Key-Word Information Systems

This paper examines how imprecision in the way humans name things might limit how well a computer can guess to what they are referring. People were asked to name things in a variety of domains: instructions for text-editing operations, index words for cooking recipes, categories for “want ads,” and descriptions of common objects. We found that random pairs of people used the same word for an object only 10 to 20 percent of the time. But we also found that hit rates could be increased threefold by using norms on naming to pick optimal names, by recognizing as many of the users' various words as possible, and by allowing the user and the system several guesses in trying to hit upon the desired target.