Language Research Sponsored by ONR

In contrast to DARPA, ONR has not had a defined program that is focused exclusively on language research. However, there have been some clusters of projects concerned with language that have been supported within the Cognitive Science Program and its prior incarnation, the Personnel and Training Research Program. The ONR program is a basic research program, but a mission-oriented one which seeks to generate results that will prove applicable to significant Navy applications. Traditionally, as the earlier name indicates, those applications were sought primarily in training; more recently, the targets have been broadened to include human factors applications, especially in human-system interaction. Originally, the program was a psychological research program, but it evolved into a cognitive science program, the first government research program to be so labeled. The ONR Cognitive Science Program emphasizes the use of AI techniques to model human cognitive performance, and it has also emphasized the special sub-field of AI concerned with artificially intelligent computerized instruction or tutoring systems (ICAI or ITS). Thus, the ONR Cognitive Science Program also contrasts with the DARPA Human Language Technology Program in being concerned with understanding how human language actually is produced and processed by humans. (There is also an AI program within the Computer Science Division of ONR.) The management style of the program might be best described as dvnamic coherence. That is, a degree of coherence or focus is necessary in order to enhance the likeliho(x:l of identifiable impact on application areas, but the focal clusters do evolve over time, partially in response to promising proposals that are received. New proposals are judged partially by the extent to which they cohere with and enhance the current portfolio of projects, not merely on isolated merit and inclusion in the broad area of cognitive science. Two salient clusters of language-related research that emerged have been the currently active emphasis on tutorial discourse and an eadier interest in improving the readability of instructional texts and documentation. Both have had significant natural language AI aspects, but not all projects have involved computation.