Advanced AI Text-to-Speech System Developed for Yoruba Language

Researchers have introduced TTSYoruba, a sophisticated rule-based concatenative diphone speech synthesizer specifically designed for the Yoruba language. This innovative system is already operational online as part of the YorubaName.com open dictionary, demonstrating its practical application in preserving and promoting African linguistic heritage. The development highlights a significant step forward in making digital tools more inclusive for low-resource languages.

The TTSYoruba system processes tone-marked Yoruba text inputs to generate audio output. Its core functionality relies on a meticulously hand-crafted phonological rule system applied to an extensive inventory of 651 diphone units. These units encompass five distinct tonal variants for every consonant-vowel combination in Yoruba, reflecting the language's complex tonal nature which is crucial for meaning.

The paper details the system's intricate phonological architecture, including its precise logic for tonal file selection and its solution to the challenging three-way nasal disambiguation problem (distinguishing oral /n/, nasalized vowels, and syllabic nasals). A notable orthographic contribution is the adoption of the caron and circumflex symbols as standard single-vowel contour tone markers, integrating them into the TTS normalization pipeline and the WriteYoruba keyboard input tool to standardize written representation of tones.

Performance evaluation involved a listener study with 50 participants, yielding detailed Mean Opinion Scores (MOS) that validate the system's effectiveness. This research not only advances text-to-speech technology but also provides a vital resource for the Yoruba-speaking population, spanning countries like Nigeria, Benin, and Togo, by enhancing digital accessibility and supporting language preservation efforts.