Cockpit Natural Language Study - Data Collection and Initial Data Analysis

摘要

This report describes an initial attempt to study language used by fighter pilots. To this end, a methodology was adopted which entailed the use of a mission scenario, tape recording the experimental sessions, and analyzing the data according to various statistical and tabular methods. Fifty-four pilots from six organizations participated in the study and generated 656 unique words, abbreviations, and acronyms. Since only the mission related commands were analyzed for this report, this number does not represent accurately the total size of the vocabulary collected in the study. However, it is the first compilation of a user-oriented voice application. The data base created in this study is meant to be a foundation for additional analyses. The statistical tables can provide researchers with sufficient descriptive material to define and conduct studies focusing on particular issues and hypotheses. Some of the major conclusions are: 1) the data gathering method and statistical/tabulation procedures are appropriate for this type of knowledge engineering; 2) as measured by vocabulary and rating scores, the sample was homogeneous with respect to differences among bases, aircraft, and pilots; 3) variations in vocabulary within situations are due to the pilots; 4) variations in vocabulary across situations are due both to pilots and to differing task demands. Keywords: Avionics, Speech technology, Natural language.