The diploma thesis addresses the development of a grapheme-to-phoneme
converter for the Slovenian language, named Govornik, which enables local
data processing, security, and stores recordings of texts that can be returned
upon repeated requests without reprocessing. In the implementation, we
used open tools and resources: the Sloleks 2.0 word-form lexicon, from which
we obtained words and their corresponding pronunciations; the Obeliks language
annotator, which helps in finding the correct pronunciation for each
word; and the open-source speech synthesizer Mary TTS, which performs the
task of converting the phonemic transcription into sound. For Mary TTS,
we built six different voices from which the user can choose. All components
were integrated with our custom grapheme-to-phoneme converter into a complete
software solution with an easily accessible API interface. On this basis,
we also developed user applications for different environments that enable
simple use of Govornik. We created end-user applications for Windows and
Android operating systems, which allow the use of this API in a user-friendly
way, an extension for Home Assistant, which enables Slovenian synthesized
speech in smart home environments, as well as an Android interface that
allows Govornik to be used in other Android applications that rely on speech
synthesis. Govornik is used in various projects and also enables communication
for people who are unable to speak due to disability or illness.
|
