2003 Conference Proceedings

Go to previous article 
Go to next article 
Return to 2003 Table of Contents 


Gregory W. Lesher, Ph.D.
Enkidu Research, Inc.
247 Pine Hill Road
Spencerport, NY 14559
Phone: 585-352-0507
Fax: 585-352-0508
Email: lesher@enkidu.net

Bryan J. Moulton
Enkidu Research, Inc.
Email: moulton@enkidu.net

Gerard J. Rinkus, Ph.D.
Enkidu Research, Inc.
Email: rinkus@enkidu.net

D. Jeffery Higginbotham, Ph.D.
Department of Communicative Disorders and Sciences
University at Buffalo
122 Cary Hall, 3435 Main Street
Buffalo, NY 14214-3005
Phone: 716-829-2797 ext. 601
Email: cdsjeff@buffalo.edu


There has recently been an increased focus on the electronic logging of augmented communication for subsequent clinical analysis (Higginbotham, Lesher, Moulton, Rinkus, 2002; Romich & Hill, 1999). This data can be used to derive quantitative measures of communicative competency (such as communication and error rates) as well as to assess the suitability of specific interfaces and augmentative techniques for an individual communicator. Under the auspices of the Rehabilitation Engineering Research Center on Communication Enhancement (the AAC-RERC), our research group has previously developed a universal logging format for electronic data recording, as well as a free analysis tool (called ACQUA) that can be used with compliant logfiles (Higginbotham et al., 2002; Lesher, Moulton, Rinkus, & Higginbotham, 2000).

The same logging and analysis tools used for clinical studies with individuals can also be applied to simulation studies designed to quickly determine the efficacy of various augmentative techniques and interfaces. In such investigations, a computer "emulates" a idealized user, automatically generating a representative message using a specific interface while recording each selection for later analysis. Over the past 5 years, our research team has used in-house tools to perform emulations of scanning interfaces, word prediction, abbreviation expansion, and a variety of other accelerative methods. These studies were designed to reveal general efficacy trends, but the same simulation procedure can be applied to specific individuals in a clinical setting. For example, knowing beforehand the vocabulary of an individual who uses scanning to spell out messages, one could rapidly assess the efficiency of various character layouts without requiring the person to laboriously test each layout him or herself.


Working with the AAC-RERC, we have developed a prototype emulation tool that provides researchers and clinicians with the opportunity to assess the performance of various augmentative interfaces. The logfiles produced by the emulation tool comply with the AAC-RERC standard, which means that they can be analyzed by ACQUA to derive such measures as keystroke efficiency, keystroke savings, button usage counts, and page frequencies. The emulation tool employs a simplified "user model", providing estimates of the maximum communication rates and selection efficiencies achievable with specific interfaces.

The emulation tool is based on Enkidu Research's Portable IMPACT Demo, Windows PC software that includes all of the features of Enkidu's commercial communication aids. These features includes word prediction, scanning, abbreviation expansion, and a broad range of other augmentative techniques. By adding emulation capabilities to an existing program, rather than building an emulation system from the ground up, we were able to produce a useful tool without requiring an inordinate development effort. The emulation tool can be freely downloaded from the AAC-RERC website (www.aac-rerc.org) and from www.enkidu.net/. The ACQUA analysis tool can likewise be found at the AAC-RERC site and at www.enkidu.net/acqua.html.


The emulation procedure is straightforward. After starting the emulation software, the operator loads a set of interface pages and configures the augmentative features as if the system were being set up for a user. The operator then turns on logging to record the desired data (for example, a simple time-stamped output), and opens a dialog window to configure the emulation. Configuration consists of defining a testing text that will be enerated during the emulation and setting parameters to determine what type of emulation will be performed (see details below). Once the emulator has been configured, the system takes over and automatically generates the testing text, logging each intermediate step in the process. The logfile produced by the emulation tool can then be analyzed using ACQUA.

Any raw text (.txt) file of up to ten thousand words can be generated by the emulator. Several sample texts are provided with the program - including the testing texts utilized in many of our early studies (Lesher et al., 1998). It is important to remember, however, that the most accurate measures will result from emulating a text sample that is representative of the messages generated by the specific augmented communicator under consideration.

By default, the emulation tool will attempt to generate the reference text as quickly as possible. This may involve taking some "shortcuts" that don't strictly match every unique action taken by a user when generating the same text, provided that these shortcuts don't change the resulting statistics. For example, when emulating a scanning system there may be no need to simulate each and every scan focus movement if the net result of these movements can be deduced mathematically. Additionally, there is often no need to update the screen or message window during emulation, since these functions will not impact the logfile. The default emulation mode - which we refer to as "Computer" mode - reduces overhead by eliminating spurious processes, resulting in very fast operation.

For some emulation studies, the operator will want to slow down the system to provide instructional or diagnostic feedback. In "Android" mode, the emulator meticulously recreates every action performed by the human user while generating the reference text. Additionally, the system shows all screen updates related to the individual emulated actions. Android emulation can be set to operate in real-time mode, in which case the system slowly generates the text like a player piano performing a waltz. Alternatively, emulation can be decelerated to provide a slow-motion version of message production, or accelerated to provide a high-speed version. It is important to note that regardless of the generation speed, emulation can be configured to record realistic time values in the logfile, so that communication rate measures can still be generated by ACQUA.

For the time being, our tool is restricted to emulating only idealized users who take the same amount of time for each action and never make any mistakes. In the case of automatic (single-switch) scanning, this approximation yields a reasonable estimate of the maximum achievable communication rate, since the fixed scanning delays account for the majority of the total duration of message production. For direct selection, communication rates are more roughly approximated using an operator-defined average action time. We have thus far ignored the more complex human factors that play a large part in determining communication rate, but recognize that modeling these elements represents an important next step in development of our emulation tool.


We have constructed AAC emulation software that can automatically generate a reference text and record the associated logging data. The quantitative measures that can be derived from the emulated data can be extremely useful in pinpointing the best interface in terms of efficiency (amount of output for each user action). For example, we used a predecessor of this tool to demonstrate the counterintuitive fact that character prediction can be more efficient than word prediction in many instances (Lesher, Moulton, & Higginbotham, 1998).

Comparing the efficiencies of various interfaces can be a valuable clinical tool, but communication rate doesn't correlate perfectly with efficiency. While the emulation tool can provide rudimentary estimates of communication rate, it would clearly benefit from a more sophisticated user model that took into account motor limitations, cognitive loads, and other factors. We plan to enhance this aspect of our simulator in the future, working from user models proposed by earlier researchers (Levine & Goodenough-Trepagnier, 1990).


Higginbotham, D.J., Lesher, G.W., Moulton, B.J., & Rinkus, G.J. (2002). Automated data logging in augmentative communication. In Winters, Robinson, Simpson & Vanderheiden, (Eds.), Emerging and Accessible Telecommunications, Information and Healthcare Technologies. Arlington: RESNA Press.

Lesher, G.W., Moulton, B.J., & Higginbotham, D.J. (1998). Techniques for augmenting scanning communication. Augmentative and Alternative Communication, 14, 81-101.

Lesher, G.W., Rinkus, G.J., Moulton, B.J., & Higginbotham, D.J. (2000). Logging and analysis of augmentative communication. Proceedings of the RESNA 2000 Annual Conference, 82-85, Arlington, VA: RESNA Press.

Levine, S. & Goodenough-Trepagnier, C. (1990). Customised text entry devices for motor-impaired users. Applied Ergonomics, 21, 55-62.

Romich, B.A. & Hill, K.J. (1999). AAC communication rate measurement: Tools and methods for clinical use. In Proceedings of the RESNA 1999 Annual Conference, 58-60, Arlington, VA: RESNA Press.


The authors wish to acknowledge support from the National Institute on Disability and Rehabilitation Research of the U.S. Department of Education under grant #H133E980026. The opinions expressed are those of the authors and do not necessarily reflect those of the supporting agency.

Go to previous article 
Go to next article 
Return to 2003 Table of Contents 
Return to Table of Proceedings

Reprinted with author(s) permission. Author(s) retain copyright.