The software is a set of Java application programming interfaces (APIs) that lets developers create what IBM is calling a transcription portlet, which transcribes words spoken through a telephone or other voice application. The idea is to let people dictate information through a phone line, to have that information transcribed into text and then to have Internet-based applications act upon the dictated text.
"We are taking the portal, which is written for a browser and (a) desktop environment, and extending it to a new set of devices," said Gene Cox, director of mobility solutions for IBM Pervasive Computing. "That means you can access it with personal digital assistants (PDAs), cell phones and other devices. We're working on new kinds of interfaces to extend it to the mobile, wireless environment."
Examples of how the application can be used within IBM's WebSphere Voice Application Access infrastructure include creating user accounts, registering e-mail accounts with which to send transcriptions, managing the transcriptions and correcting them.
The transcription portlet, released April 17 and available as a free, trial download, is based on the World Wide Web Consortium (W3C) standard called VoiceXML, an Extensible Markup Language dialect.
IBM's Voice Application Access technology already let IBM's personalized business portals respond to some voice commands. But with the introduction of the transcription portlet on IBM's AlphaWorks site for technologies under development, IBM now claims that its vocabulary is in the "hundreds of thousands" of words.
The second edition of the Oxford English Dictionary, by contrast, contains full entries for 171,476 non-obsolete words.