W3C Expands Support for Speech Synthesis of World Languages
W3C took steps to broaden support for the world's languages in voice applications on the Web. This First Public Working Draft of Speech Synthesis Markup Language (SSML) 1.1 incorporates features and feedback from SSML Workshops held in Beijing, China and Heraklion, Greece. On January 13-14, 2007, W3C is conducting a third Workshop on SSML, hosted by Bhrigus Software and the International Institute of Information Technology (IIIT) in Hyderabad, India. This Workshop promises more review and contributions to SSML, part of W3C's Speech Interface Framework, a suite of specifications for building voice applications on the Web.
It is forecast that within three years, the World Wide Web will contain significantly more content from Chinese and Indian language families, among others. In many of the regions where these languages are spoken, people can access the Web mo through a less expensive mobile handset than through a desktop computer. Today the world has more than 10 times as many cell phones as Internet-connected personal computers. With an improved SSML, people worldwide will have an increased ability to listen to synthesized speech through mobile phones, desktop computers, and other devices, extending the reach of computation and information delivery.
SSML 1.1 improves on W3C's SSML 1.0 Recommendation by adding support for more conventions and practices of the world's languages. One new feature helps to disambiguate word boundaries in languages that do not use whitespace as a word boundary, including Chinese, Thai, and Japanese. SSML 1.1 allows references to language-specific pronunciation alphabets. It clarifies the relationship between the author's specified speaking voice and the language being spoken. It provides finer-grained control over lexicon activation and entry usage.
In addition, SSML 1.1 provides features to better integrate with existing and upcoming Speech Interface Framework specifications.
The third SSML Workshop brings together experts from India, Pakistan, and other countries to identify and prioritize requirements for SSML extensions and additions that will improve its use for rendering non-English languages including (but not limited to) Arabic, Hebrew, and the Indian languages Telugu, Hindi, Bengali, Punjabi, Gujarthi, and Urdu. The top priorities on the agenda are to describe new requirements, their usage scenarios, and the problems to be solved. The output from this workshop will be reviewed by the W3C Voice Browser Working Group for possible new features in SSML 1.1 and beyond.