How to teach Slovak TTS to speak (Serviko) Romani

The paper presents the first step of our attempt to enable speech driven information technologies in Romani. A short overview is given of Romani orthography, phonetic inventory, pronunciation rules, and prosody. Differences between Slovak and Romani language systems are pointed out and the resulting knowledge is used to design parts of the Romani diphone synthesizer derived from the corresponding parts of our Slovak diphone synthesizer Kempelen 1.2 with male voice Milan and Kempelen 1.4 with female voice Anna. The first listening experiments indicate that the proposed procedure porting an existing Slovak speech synthesizer to Romani is effective and represents for us the fastest and cheapest way of creating a well understandable basic level synthesizer in Romani. This synthesizer can be immediately used in the design of speech driven information systems in Romani. But the main achievement of this work is that we have acquired skills in processing new language, we have found its main peculiarities, and have prepared the basic speech processing background needed for further development of much more comprehensive speech technology applications like paedagogical tools and information systems using unit-selection synthesis and automatic speech recognition in Romani.