Third Workshop on African Language Technology (AfLaT 2011) - Report

Towards Standardizing Evaluation Test Sets for Compound Analysers

Liaan L. Fourie

Liaan Fourie and colleagues describe best practices for the evaluation of compound analyzers for Afrikaans (and other languages). Different options are explored and quantitatively evaluated. The presentation provides pointers for related research efforts on compound analyzers.
[Full Abstract]

A Number to Yorùbá Text Transcription System

Tunji Odejobi

The pronunciation of numbers presents a serious bottleneck for text-to-speech systems. Tunji Odejobi and colleagues provide an exhaustive overview of different possible pronunciation patterns for Yoruba and a finite-state-technique to represent them as tree structures.
[Full Abstract]

Modeling Improved Syllabification Algorithm for Amharic

Nirayo Hailu

The syllable structure of Amharic is described, in the context of stress assignment for speech synthesis. Amharic-specific morphological features, such as gemination and epenthesis are handled in an expansive framework that is able to syllabify words with a state-of-the-art accuracy of over 98%.
[Full Abstract]