Prosodic Modeling for Amharic
No Thumbnail Available
Date
2008-04
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
During daily speech communications, we systematically employ variations in pitch,
loudness, phrasing and duration of spoken segments to express our intentions, attitudes
and assumptions. These cues are continuously picked by our listeners and our intentions
are communicated. This phenomenon is what is termed as prosody. Various languages
have been modeled using a prosodic modeling framework and have been used to varying
degrees of success in applications like speech synthesis systems. The prosodic features
of speech are largely dependent on the language, thus it is imperative that prosodic
models be developed and tested for each language to apply them in (language specific)
speech synthesis applications. In this study the prosodic nature of Amharic is
investigated, and based on the investigation the prosodic cues of Amharic are identified
and an Amharic prosodic model based on the ToBI framework is developed. This
hypothesized model is then be subjected to a rigorous test to determine whether it results
in improvement of Amharic synthetic speech in terms of naturalness and intelligibility.
Description
Keywords
Amharic