If you have the Dzongkha National Language Processing app on your phone, you will now be able to translate Dzongkha to English and vice versa. Not only that, the app can also convert Dzongkha audio into text and Dzongkha text into audio. The system which is a first of its kind and available as both app and web-based system was launched today in Thimphu. The system was initiated under the Digital Drukyul Flagship Programme in 2019.
The system has three features namely the Dzongkha NMT or Neural Machine Translation which translates Dzongkha and English texts and Dzongkha ASR or Automatic Speech Recognition which transcribes Dzongkha speech into text. The third feature is Dzongkha TTS or Text to Speech synthesis which transcribes Dzongkha text into speech.
Until now, the technology for Dzongkha has been limited to input, storage and display. However, now with the help of the Dzongkha National Language Processing or NLP System, it is expected to enhance the usage and processing of Dzongkha using technology.
According to the Department of Culture and Dzongkha Development, machine translation is important for breaking language barriers and providing equal access to information and services. It is also required in making the language easy to use.
“With the change in time, technology is very important. So, it is important to digitise Dzongkha to preserve it. Globally, all people are now used to using various technologies. If we could just set this introduction as a foundation then it can be a motivation in the future to develop more advanced Dzongkha software,” said Namgay Thinley, Chief Programme Officer.
The College of Science and Technology which developed the system will also continue to assist the project to upgrade its versions.
“This type of tool especially that involves Dzongkha is very rare, whereas in other countries, there are so many automated translation tools. We are happy that the first version has come out successfully. Hereafter, we will not leave the project just like that but try to upgrade the versions which will include grammar and spelling checker and OCR as well,” said Pema Galey, Associate Lecturer.
The system will further be enhanced to deliver accurate results as the dataset of the system has to be widened.
Currently, as the dataset incorporated is not that large, the translation accuracy falls between 30 to 50 per cent.
The app is available on android phones currently, however, it will also be made available on iOS devices soon.