Google has established a cooperation with UGM to develop Text-To-Speech (TTS) for Javanese language in their project called, Project Unison. In this project, UGM will facilitate the collection of Javanese utterance for a database needed for the application.
“It’s very important for us to work with universities because data collection process is very difficult to do, particularly for regional languages,” said Senior Program Manager of Google Research for Low Resource Languages, Linne Ha, in the Seminar Digitalisation of Javanese Language with Google ASR and TTS on Wednesday (4/5) in the Auditorium of Faculty of Cultural Sciences UGM.
Project Unison is a project started by Google in 2015 to explore innovative application Text-To-Speech aimed for low resource languages. In Indonesia, Javanese is the first to be digitalised as the speakers is the largest. Even so, Linne added, the dialect of Javanese that varies poses a challenge for the app maker. “The variation of Javanese dialect is a challenge, but we try to collect as many data as possible to build sound identification that is able to grasp the various dialects,” he said.
The data collection does not use the normal practice with voice talents and professional studio but by recording voices of speakers under various environments to catch the variety of dialects.
The cooperation is a form of UGM’s participation in sustaining local cultures. “Presently, not many people speak Javanese. (Let’s prevent the possible situation) when in the next few years, Javanese language is known only by academics,” said Head of UGM Data and Security Centre in the Information System and Resource Directorate, Dedi Eko Yunanto Priyadi, S.T.
He emphasised the importance of the cooperation as the start of Google research in Indonesia. “This cooperation will be an example for the digitalisation of other regional languages in Indonesia. If all is well, this will open the way for the app to be developed for other regional languages,” he said.