Home > News & Events > MITRE Publications > The Edge >

Spoken Language Case Study

Underwater HCI

Last year MITRE and the Johns Hopkins University Applied Physics Laboratory joined forces to transition recent developments in speech interfaces to the submarine community at the behest of the Navy's Advanced Systems and Technology Office (ASTO) and the Naval Sea Systems Command (PMS425). The collaborators are evaluating new interfaces in parallel with the Navy's Advanced Processing Build program evaluations of new sonar and signal processing algorithms.

The driving factor is the back-end application, a data fusion sonar repeater for the officer of the deck (OOD). Human factors and system issues are both considered: who uses the application, how and where will it be used, hardware and software constraints, and whether the interface is a retrofit or part of a new development path are all relevant issues in developing an interface.

The OOD display is used in various operating environments. Ambient noise levels are different at anchor and under way. This raises the most critical challenge from the standpoint of speech recognition technology -- robust recognition of speech in background noise. MITRE is evaluating commercial and research speech recognition systems for robustness and will soon start to look at other components such as natural language and dialog systems that are in the research domain. In addition to the individual components, the research community has recently started looking at architectures and standards (see Communicator article on page six).

MITRE's goal is to provide additional functionality to the standard graphical user interface, while not supplanting the keyboard. However, the additional functionality introduces a hardware constraint of its own, in that it requires a microphone. Most speech recognizers require head-mounted, noise-canceling microphones. Submarine control room operators, however, must use wireless devices, and a minimally obtrusive microphone is desirable.

The back end will have multiple users. Whereas human operators can prioritize input from multiple users based on an individual's role in the command hierarchy, software systems cannot. We therefore constrain input to one user at a time.

HCI for Training Simulations

MITRE has been at the nexus of several efforts that have resulted in integrated speech interfaces with training simulations. Spoken interaction provides an increased span of control to alleviate the burden placed on the human-in-the-loop in distributed simulations and ultimately can result in allowing the user trainee to interact directly with the simulation.

MITRE was part of the team that successfully integrated and deployed the SRI International CommandTalk spoken-language interface with the joint Semi-Automated Forces (SAFs) in the Synthetic Theater of War (STOW-97) interoperability exercise. This year, the Army's Simulation, Training and Instrumentation Command (STRICOM) asked MITRE to transition CommandTalk to the Dismounted Infantry SAF (DISAF) component of the Modular SAF (ModSAF) 5.0 application. CommandTalk was demonstrated as a spoken language interface to the ModSAF 5.0 application by enabling DISAF entity, unit, and execution matrix setup, and command and control (C2) of DISAF computer-generated forces.

All Talking, All Drawing

MITRE is currently working with both the Army's Communications-Electronics Command (CECOM) and the Defense Advanced Research Projects Agency (DARPA) to transition spoken-language interfaces into battlefield environment systems.

As a proof of principle experiment, MITRE transitioned a spoken language interface called QuickSet, developed at the Oregon Graduate Institute, to a Web-based next-generation prototype of the Army's Maneuver Control System (MCS). Quickset's unique characteristic is that it integrates both speech and drawing, allowing users to set up forces on a map with commands such as "Set fourth platoon here," in which " here" is indicated by drawing on a map. MITRE is setting Quickset on a migration path with the DARPA Communicator standards by making it compliant with the Communicator architecture. The end result will be an exercise in portability to Communicator standards for DARPA, and a vision of the potential impact of spoken language technology for the Army.