Speech Technology for Persons With Disabilities: Are We Breaking Down Barriers or Creating New Ones?
Julie is quadriplegic, and the most effective way for her to interact with technology is by voice. In her work as a paralegal, she uses speech recognition software to operate the computer and to dictate documents. At home, she uses a speech-operated environmental control system to operate lights, TV, stereo, and other household controls and devices. Julie credits advances in speech technology with allowing her to have a career and to live independently. She has been using speech recognition technology since the early 1990s, when the only available products required discrete speech (pausing between each word) and had very limited vocabularies. Today, thanks to an expanding mainstream market for speech recognition, she uses software that allows her to speak more naturally, has much better accuracy, and costs far less than the product with which she started.
Steven is blind and works as a software developer. Because he can use the standard keyboard, he doesn't use speech recognition technology for input, but he depends on speech synthesis to access his computer's output. Although he's a proficient computer user, he currently faces many challenges, since software, Web sites, and other information technologies are often presented for a visual audience and are not accessible to those using an audible interface. He hopes and advocates for a day when speech becomes a mainstream modality for interfacing with technology, forcing developers to create products and content that can be perceived both audibly and visually.
For Julie, Steven, and millions of others, the mainstreaming of speech technology represents a significant improvement in the accessibility of the world around them. However, while speech technology is breaking down barriers, it also has the potential to erect new ones. If a device or system is operable via speech input, it must also be operable by individuals who are unable to speak clearly. If a device or system provides speech output, it must also provide output via a medium that is perceivable by individuals who are unable to hear.
A well-designed, accessible solution will employ the philosophy of universal design. Universal design is the process of creating products that are accessible to people with a wide range of abilities and characteristics. The concept originated in the field of architecture, when building designers first clearly stated and implemented principles intended to make facilities accessible to the broadest possible number of people. The concept can be extended to information technology (IT): An IT system that is developed according to universal design principles is one in which the same transactions or interactions can occur via multiple media (e.g., phone and Web) and multiple modalities (e.g., voice input complemented with touchtone input, audible output complimented with text-based output.)
At SpeechTEK 2004, nine teams participated in the Speech Solutions Challenge II. Teams were given six hours to build a specified speech application from scratch. Then, SpeechTEK attendees were given an opportunity to sample and evaluate the finished products via a toll-free telephone line. For my evaluation, I first simulated a user with compromised speech, and then simulated a deaf user interacting with the application through a telephone relay operator. With my simulated disabilities, I successfully completed a transaction using three of the nine applications. The applications that worked for me worked for two reasons:
They provided true multimodal functionality (i.e., I could use the keypad if my speech wasn't clear). They provided me with sufficient time to respond to questions, and if I still took too long, they provided me with the means to essentially pause the application (e.g., "If you need a minute, say, 'Wait a minute,' or press the number 1 on the keypad"), as well as the means to resume when ready.
These same features could of course benefit users without disabilities as well. Although speech technology continues to improve, it is always likely to have some degree of difficulty understanding human speech when there is significant background noise or when the user speaks in a heavy accent. Similarly, users with and without disabilities might for various reasons need more time to respond. Universal design expands the scope of users, with and without disabilities, who can successfully utilize an application.
Universal design is also necessary for legal reasons. Section 255 of the Telecommunications Act of 1996 requires manufacturers of telecommunications equipment to ensure that their equipment is designed, developed, and fabricated to be accessible to and usable by individuals with disabilities. According to Federal Communications Commission (FCC) rule, voicemail and interactive voice response (IVR) systems are covered by Section 255. Also, 1998 amendments to Section 508 of the Rehabilitation Act require accessibility of electronic and information technology (EandIT) procured, maintained, or used by the federal government. According to Section 508 and its accompanying standards, EandIT includes software applications, Web-based intranet and Internet information and applications, and telecommunications products, among other technologies. Also, the Americans with Disabilities Act (ADA) of 1990 prohibits discrimination on the basis of disability by private employers (Title I), state and local governments (Title II), and places of public accommodation (Title III). In June 2002, the U.S. Eleventh Circuit Court of Appeals ruled that Valleycrest Productions had violated the ADA because individuals with hearing and upper-body mobility impairments were unable to complete the automated telephone screening process for the TV show Who Wants to Be a Millionaire? The system timed out too rapidly for the plaintiffs to respond and provided no opportunity for accommodation.
The Alliance for Telecommunications Industry Solutions (ATIS), an organization that develops and promotes standards for the communications industry, has assembled a forum to investigate and document accessibility issues, particularly those affecting IVR systems and services. The IVR Accessibility Forum has responded by assembling a wealth of information. The forum's Web site (http://www.atis.org/ivr) includes a Disability Implications Matrix (which can be viewed by specific application, user interface element, or disability), an IVR Accessibility Checklist, and extensive information about each of the three laws described in the preceding paragraph.
Regardless of whether the justification is legal compliance or expansion of target audience, the future looks bright as the speech industry moves forward, increasingly expanding speech technology into the societal mainstream. If industry players embrace universal design, all users - with and without disabilities - will finally be able to participate in our technology-enhanced world.
Terry Thompson is senior computer specialist with the University of Washington (UW), where he divides his time between AccessIT (the National Center on Accessible Information Technology in Education), and DO-IT (Disabilities, Opportunities, Internetworking and Technology). He has presented at numerous national conferences, seminars and workshops on the topic of technology accessibility, and has consulted with government as well as K-12 and postsecondary education entities on technology access issues.