The effects of a system-imposed grammatical restriction on interactive natural language dialogue

Autor: Hendler, James, Michaelis, Paul Roller
Zdroj: Research on Language and Social Interaction; 1984, Vol. 17 Issue: 2 p127-143, 17p
Abstrakt: All interactive natural language computer systems require users to limit themselves to specific subsets of English. Although such language limitations reduce the computer's processing burden, the effects on users are less clearly understood. This experiment examined the effects of a system-imposed grammatical limitation on two-person teams who cooperatively solved problems.Twenty-four subjects were randomly assigned to twelve two-person teams. Each team solved three problems, which were presented in a completely counterbalanced order. Team members communicated by typing messages on computer terminals. Six of the teams were in an unrestricted condition. However, the allowable grammar for the remaining six teams was highly restricted.Dependent measures included time-to-solution, problem solving, accuracy, and various measures of verbal output. Significant differences were found only for the measures of time-to-solution, words per sentence, and, in the restricted condition, the number of attempted “ungrammatical”; messages.The unrestricted teams took an average of 21.5 minutes to solve each problem, versus 30.55 minutes for the restricted; F(l, 10) = 5.74, P < 0.05. There was also a significant interaction of Grammar with Sessions, F(2,20) = 3.62, P < 0.05: averaged across the first session, the restricted teams were slower (46.33 versus 25.25 minutes), however, their performance on the second and third sessions differed insignificantly from that of the unrestricted teams. The restricted teams' improved performance is also illustrated by a significant decrease in the number of ungrammatical messages across sessions, averaging 9.33 per team in the first session, 3.83 in the second, and 2.83 in the third.The main effect of Grammar was also statistically significant for the dependent measure of words per sentence. The unrestricted teams used an average of 14.12 words per sentence, versus 9.07 for the restricted; F(l, 10) = 8.49,P < 0.01.In summary, there is strong evidence that the restricted subjects in this study quickly adapted to the grammatical restrictions that had been imposed on them.
Databáze: Supplemental Index