The University of Edinburgh -
Division of Informatics
Forrest Hill & 80 South Bridge

MSc Thesis #9532

Title:Ssml: a Markup Language for Speech Synthesis
Date: 1995
Abstract:At present, most speech synthesis systems use unconstrained text as their input, which is desirable from a human point of view, but problematic as far as machines are concerned, since the process of converting text to speech is very complex. In this thesis, the development of a Speech Synthesis Markup Language (SSML) is described. The SSML format is a compromise between human and machine needs. The input is text-based and unconstrained in its use of words, but allows a large amount of extra information to be included with the text to guide the synthesis in its performance. The text is annotated with markers to specify features such as emphasis, particular speech styles, or the beginning of new topics. A prototype version of SSML is described, with the Markup-to-Speech program written for use with the CSTR speech synthesizer. A larger set of markers for a future version of SSML is also presented.

[Search These Pages] [DAI Home Page] [Comment]