eXTRA: A Culturally Enriched Malay Text to Speech System

This paper concerns the incorporation of naturalness into Malay Text-to-Speech (TTS) systems through the addition of a culturally-localized affective component. Previous studies on emotion theories were examined to draw up assumptions about emotions. These studies also include the findings from observations by anthropologists and researchers on culturalspecific emotions, particularly, the Malay culture. These findings were used to elicit the requirements for modeling affect in the TTS that conforms to the people of the Malay culture in Malaysia. The goal is to introduce a novel method for generating Malay expressive speech by embedding a localized ‘emotion layer’ called eXpressive Text Reader Automation Layer, abbreviated as eXTRA. In a pilot project, the prototype is used with Fasih, the first Malay Text-to-Speech system developed by MIMOS Berhad, which can read unrestricted Malay text in four emotions: anger, sadness, happiness and fear. In this paper however, concentration is given to the first two emotions. eXTRA is evaluated through open perception tests by both native and non-native listeners. The results show more than sixty percent of recognition rate, which confirmed the satisfactory performance of the approaches. 12