... using.[*]
One might imagine that the industry ``standard'', Microsoft's SAPI, has solved this portability problem, but this is in fact not the case: only minimal requirements of compliance are required for a system to be able to claim that it is SAPI compliant, and thus there is no guarantee that the same set of tag specifications will yield comparable results when used with two distinct ``SAPI-compliant'' systems.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... text[*]
ISO639 defines two-letter language codes for around 140 languages. For example, ``de'' identifies German.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... simple.[*]
It has often been suggested that rather than use a speech specific markup such as SABLE, it would be better to develop mechanisms for spoken interpretation of existing tags in other visual markup languages such as HTML or Latex. This direct interpretation approach is unattractive for a number of reasons. SABLE can be used as an interface language in systems for which HTML would be inappropriate, for instance when coupling a natural language generation system with a speech synthesizer. Leading on from this, there are tags in SABLE which are speech specific, such as PRON. It would be unreasonable to expect designers of visual markup languages to incorporate such tags in their markup and even if they did it would be unreasonable of authors of documents in that markup to specify PRON tags as they would seem irrelevant for most purposes. Finally, SABLE is intended as a standard synthesizer interface - if a speech specific markup language was not used, a separate interpreter would have to be built for each markup language for each synthesis system. To speak documents authored in other markup languages we advocate building markup language filters, that convert from visual markup to SABLE. That way, only one filter need be designed for each visual markup language, and builders of synthesis systems need only build an interpreter for SABLE.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Richard Sproat
1998-11-16