Contents Previous Next
This document is the technical specification for the Java Speech API Markup
Language (JSML). JSML is a platform-independent, vendor-independent format
for text input to speech synthesizers. JSML is an XML application (eXtensible
Markup Language) that enables markup of text input to speech synthesizers to
indicate document structure, control pronunciation of words, and guide the
phrasing, emphasis, pitch, speaking rate and other characteristics of speech
output. These features enable users and applications to produce speech output that
is more natural and understandable than unmarked text.
Review Comments
This version of the JSML specification is a beta draft. We are seeking public input
concerning the Java Speech API Markup Language specification. Send your
comments and suggestions to:
Please be sure to include the version number and date of the document you are
reviewing with your comments. The earlier we receive your feedback, the more
likely it will be taken into consideration for the next update.
About JSML V0.6
This revision of JSML introduces significant changes from v0.5 (a detailed list of
changes is included at the end of the preface). These changes reflect experience
with JSML usage since its first release, movements in the XML specification (on
which JSML is based), and advances made by the speech synthesis community on
text markup through the SABLE specification (which is derived, in part, from
JSML v0.5).
Several of the changes introduces with JSML in this release break compatibility
with existing JSML documents and existing JSAPI speech synthesis
implementations. While we were reluctant to break compatibility, these changes
are important to ensuring that JSML provides a robust standard that will support a
diverse set of application needs for many years, and to ensure JSML is consistent
with other applications of XML.
We recommend that speech synthesis implementations move to the latest
specification as soon as practical. If an implementation continues to support the
old specification it should provide a warning whenever an old JSML document is
detected and that support should be phased out to ensure long-term consistency of
JSAPI and JSML implementations.
Technical Scope
This specification describes a format for marked-up text input to a speech
synthesizer. However, this specification does not address the issues listed below.
These programmatic issues are covered in the Java Speech API specification
which defines the software interface for speech synthesizers as an extension to the
Java platform.
- Mechanisms for providing marked-up text to a speech synthesizer.
- Software control of the output of annotated text such as queuing, pause and
resume, and variation of pitch and speaking rate.
- Mechanisms for receiving notification of synthesis events including marker
events requested in JSML texts.
- Error handling capabilities including detection of incorrect markup.
- Vocabulary management issues such as provision of pronunciations.
Web Resources
For more information about the Java Speech API and for the latest information on
the Java Speech API Markup Language visit the Java Speech API home page:
To obtain information about other Java Media APIs visit:
Mailing Lists
Discussion lists have been set up for everyone interested in the Java Speech API,
the Java Speech API Grammar Format specification, and the Java Speech API
Markup Language. The javaspeech-announce mailing list carries important
announcements about releases and updates. The javaspeech-interest
mailing list is for open discussion of the Java Speech API and the associated
specifications. To subscribe to these mailing lists send email containing either or
both of the following lines:
subscribe javaspeech-announce YOUR NAME
subscribe javaspeech-interest YOUR NAME
to:
To send messages to the interest list, send email to:
The javaspeech-announce mailing list is moderated and it is not possible to
send email to that list.
To unsubscribe from the javaspeech-announce or javaspeech-interest
lists, send email with either of the following lines:
signoff javaspeech-announce
signoff javaspeech-interest
to:
Future Releases
The field of speech synthesis is continuing to evolve with ongoing research and
new products. As new functionality becomes widely supportable, JSML can be
extended to support that functionality. Some capabilities that may be considered
in future releases include:
- Elements of attributes to indicate intonational patterns for phrases and other
structural entities.
- An element or attribute that enable a change of language within a document.
- A new element or an extension to the "phoneme" element to specify detailed
phonetic strings with pitch and timing information.
- An element that allows audio files to be embedded within or synchronized
with text output.
Revision History
Version 0.6
- All tags and attribute names changed to lower case since this is the convention
for XML applications. Where practical, tag and attribute names have
been changed so that they are not abbreviated.
- The "
div" element subsumes the functionality of the "PARA" and "SENT"
by specifying the structural entity type as an element attribute. This provides
a smoother path for future revisions to add new structure types (e.g.
discourse steps that have special spoken characteristics).
- Implicit paragraphs are no longer supported within the "
div" element.
- Added a "
voice" element to request changes in speaking voice.
- Validation behavior is specified in conventional XML terminology.
- Added a language attribute to the top-level "
jsml" element that indicates
the language contained by the document.
- The functionality of the "
SAYAS" element in JSML v0.5 is split into a new
"phoneme" element and a modified "sayas" element.
- The "
EMP" element is "emphasis" and is not legal as an empty element.
- The "
BREAK" element is "break" and its "MSECS" attribute has changed
to the "time" attribute with a standard CSS2 value format.
- Other element and attribute name changes: "
PROS" tag changed to "prosody
"; "VOL" attribute of "prosody" changed to "volume"; "ENGID" attribute
of "engine" element changed to "name"
- The attributes of the "
prosody" element now accept descriptive values
(e.g. "fast", "slow") as alternatives to numerical values.
Version 0.5
- First public beta release of JSML.
Contents Previous Next
|
Print-friendly Version
