|Dāvis Nicmanis (Tilde, Latvia), Askars Salimbajevs (Tilde, Latvia)|
To fully enable spoken human-computer interaction, the text-to-speech (TTS) component of such a system must produce natural human-like speech and adjust the prosody according to the dialog context. While the current publicly available TTS services can produce natural-sounding speech, they usually lack emotional expressiveness. In this paper, we present an expressive speech synthesis prototype for the Latvian language. The prototype is integrated into our chatbot management system and enables bot designers to specify the stylistic information for each bot response, thus making the interaction with the chatbot more natural.