Transcoding multilingual and non-standard web content to voiceXML

Master Thesis


Permanent link to this Item
Journal Title
Link to Journal
Journal ISSN
Volume Title

University of Cape Town

Transcoding systems redesign and reformat already existing web interfaces into other formats so that they can be available to other audiences. For example, change it into audio, sign language or other medium. The bene_t of such systems is less work on meeting the needs of di_erent audiences. This thesis describes the design and the implementation details of a transcoding system called Dinaco. Dinaco is targeted at converting HTML web pages which are created using Extensible MarkupLanguage (XML) technologies to speech interfaces. The di_erentiating feature ofDinaco is that it uses separated annotations during its transcoding process, while previous transcoding systems use HTML dependent annotations. These separated annotations enable Dinaco to pre-normalize non-standard words and to generate VoiceXML interfaces which have semantics of content. The semantics help Textto-Speech (TTS) tools to read multilingual text and to do text normalization. The results from experiments indicate that pre-normalizing non-standard words and appending semantics enable Dinaco to generate VoiceXML interfaces which are more usable than those which are generated by transcoding systems which use HTML dependent annotations. The thesis uses the design of Dinaco to demonstrate how separating annotations makes it possible to write descriptions of content which cannot be written using external HTML dependent annotations and how separating annotations makes it easy to write, maintain, re-use and share annotations.

Includes abstract.

Includes bibliographical references (leaves 112-119).