pwWebSpeak, pronounced as “P W Web Speak”, is a non-visual Web browser developed by a company called the Productivity Works. Formally released in 1996, the browser evolved from an initial prototype developed by Markku Hakkinen in 1995 called WebTalk. Discovering that the name WebTalk was already in use, the browser was renamed WebSpeak, along with the prefix, pw. John De Witt provided valuable insight and assistance during the early design, and Thomas Edison State College, a distance learning institution, was an early supporter of the effort.
pwWebSpeak was unique at its time, in that it started from the premise that HTML could be directly converted in a meaningful manner to audio using synthetic speech, bypassing any visual rendering of the content. This approach was termed first order design. Using a rule base, called TLD, the Tag Language Definition, pwWebSpeak parsed HTML content to create a navigable, structured audio version of a Web page. Rules included marking HTML elements as structurally significant (meaning they became points that could be navigated to within a document, such as headings), speech and non-speech audio cues that could be inserted before and after the occurrence of any HTML element (such as audio cues to indicate a hyperlink), and changes in speech synthesizer parameters based on tags. The design of the user interface and TLD allowed for quickly adapting the way in which the browser processed and voiced HTML content. The parser and TLD were effectively language neutral, allowing TLDs to be created for arbitrary markup languages. Keyboard commands were extensive and easily customized to support new functionality.
Though the emphasis was on non-visual rendering, pwWebSpeak did display a text version of the processed HTML, including any text insertions defined in the TLD. The text display, which presented a synchronized highlighting of text being spoken, included controls for altering foreground and background colors, font, and font size. Images, though not included in the text display, could be examined via an image viewer application. The browser supported the RealAudio and Windows Media Player plug-ins, in addition to basic support for DAISY talking books in the Sigtuna Digital Audio Browser, a version of pwWebSpeak created for the Japanese Society for the Rehabilitation of Persons with Disabilities. pwWebSpeak was translated and supported in a variety of languages, including Japanese.
The browser included a speech synthesizer, called SoftVoice, and supported Microsoft Speech API 4 synthesizers, as well as several synthesizers with unique APIs for Japanese, Finnish and Italian.
pwWebSpeak was only available for the Microsoft Windows platform. A PC-DOS version was developed in 1996, but dropped in favor of Windows. During 2000, as part of a project to develop a DAISY digital talking book player for the SEGA Dreamcast, a version of pwWebSpeak was contemplated for that platform, but the lack of a speech synthesizer brought the effort to a halt.
A hybrid version, unofficially named pwIE, was in development but did not reach product stage. The pwIE version married the non-visual interface of pwWebSpeak with the visual presentation and broader support for scripting in Microsoft Internet Explorer. This hybrid version evolved into pwKiosk and later isKiosk, an application for creating accessible information kiosks based upon Web content. The most notable installation of pwKiosk was in 2000 at the Smithsonian Institution’s Museum of American History exhibit marking the 10th Anniversary of the American’s with Disabilities Act.
The core engine of pwWebSpeak was also adapted in 1998 to enable telephone-based browsing of the Web, through a product called, not surprisingly, pwTelephone. Though functional, the system encountered numerous technical challenges with telephony cards and support for the Japanese telephone network. isSound discontinued the development of pwTelephone and moved toward a VoiceXML-based application set called ActiveReading and DocsByPhone.
In addition to Markku Hakkinen, the development, documentation, and support team for pwWebspeak included Linda Dorrian, Ray Ingram, and Mickey Quenzer.
pwWebSpeak is no longer officially available, though copies of the installation software and registration codes are still occasionally found on the Web. The browser is known to run on Windows XP, and still effectively renders non-scripted HTML content. The ability of the browser to process arbitrary markup languages was used by the author in 2007 to examine how the Common Alerting Protocol could be rendered using synthetic speech.
Note: Additional information and references will be added as I dig them out of my archives. If you have specific comments or questions related to pwWebSpeak, please let me know via mhakkinen at acm.org.