Abstract
Because many Department of Defense research contracts require a
level of technical guidance and oversight outside the knowledge of the
Contracting Officer, it is necessary for research scientists or engineers,
as Contracting Officer's Technical Representatives (COTRs), to provide
this expertise in the administration of the contract. This paper presents
a case study in the application of expert systems technology to Defense
research contracting. It describes the life-cycle development of CESA, a
COTR Expert System Aid built by the Navy Center for Applied Research in
Artificial Intelligence (NCARAI) at the Naval Research Laboratory (NRL) to
provide assistance to the COTR community in the pre-award phase of
research contract administration. In particular, it discusses the
selection of a specific contracting problem domain, describes the
prototype development processes, discusses system testing and evaluation,
and describes prototype maintenance during a projected period of extensive
field testing. The paper also identifies factors contributing to the
success of CESA, and concludes with a discussion of research issues
relating to further CESA development.
Abstract
An immersive language learning environment undertakes to engage
the student in a two-medium communication process: a conversation
supplemented by graphical interaction in an ordinary scene on the computer
screen. The fundamental rationale for such a system is that it promotes
language learning by enabling the student to use the new language, not
analyze or translate it. In this paper, we examine two constellations of
issues that arise in trying to provide computer-based language immersion,
issues concerning discourse and issues of tutorial strategy, and consider
how to deal with their apparently conflicting demands.
Abstract
An intelligent, video-oriented instructional system for aircrew
coordination training (ACT), is being devised and implemented as a
generalization of exercises for ACT instructors, developed by our
colleagues at the Naval Training Systems Center. Work on this VIS/ACT
system currently has two foci. One is an interface for knowledge capture
from an expert and a simplified version of it for accepting instructor
trainee (IT) responses to videos of flight simulation episodes. These
latter responses take roughly the same form as the ones the IT would be
expected to make in actual practice as a real instructor. The second
thrust is the diagnostic component of the delivered system. It uses rules
and computations, expressing both episodic and general knowledge acquired
from the expert, to evaluate various aspects of the IT's performance. The
resulting critique will then permit the pedagogical component to guide the
IT in further study of the video episode. Some possible choices of what
to do are to report the most serious performance flaws, give explanations
for the expert's choices, and present or review relevant portions of the
current and other episodes. To broaden the potential applicability of the
system, we have attempted from the outset to generalize the problem.
Abstract
BaRT is an inference engine which has been developed to aid in
classification problem solving. This inference engine uses Bayesian
reasoning and can handle problems associated with incomplete and uncertain
evidence. It has successfully been used to perform ship classification.
This manual describes how to load the BaRT program and how to use all of
the available commands. This manual also provides some theoretical
background and some implementation details concerning BaRT.
Abstract
The Navy Center for Applied Research in Artificial Intelligence
(NCARAI) at the Naval Research Laboratory conducts applied research in
artificial intelligence (AI) aimed at demonstrating the applicability and
effectiveness of AI methods to practical problems. Several systems
developed at NCARAI have gone through the research phase, reached
maturity, and are available for distribution to universities, industry and
government laboratories. A surprisingly large fraction of these efforts
was required to address means to transition artificial intelligence
technology into service. The experiences gained during domain selection,
system development, packaging, testing and distribution may serve to
provide insights for others contemplating similar packages.
Intelligent M4 Systems
Machine Learning
Neural Networks
Sensor-Based Systems
Back to NCARAI Library