Local techies are building a voice-based digital infrastructure that will allow Kinyarwanda speakers to perform various tasks on the Internet and carry out other information technology-based tasks in the local language.

The platform is part of Mozilla (a global Internet company) project called “common voices” that seeks to create voice systems that will be used to develop computer-based solutions using languages spoken around the world.

Now, a group of local developers under the Digital Umuganda project are looking for thousands of digitally recorded voices in Kinyarwanda that will be used to build a data set on Mozilla platform and enable the system to recognise Kinyarwanda among other languages.

Last week, developers held a Digital Umuganda exercise to collect data voices from volunteers who donated their voices to read Kinyarwanda texts, which are currently available on the platform.

“We are basically building an infrastructure,” said Audace Niyonkuru, managing director of the Digital Umuganda project, adding, “Software developers, private companies, NGOs and anybody who is interested will create a solution on top of that infrastructure.”

Developers said data voice collection is like teaching the computer Kinyarwanda, so that it can recognise the language that is spoken by about 40 million people in the Great Lakes Region in Africa.

“It will offer Rwandans access and services in Kinyarwanda, but there are also other benefits like making technology more inclusive for people with disabilities and those who cannot speak foreign languages,” said Mr Niyonkuru.

Using voice to perform tasks

Voice technologies are growing around the world to allow people to use their voices to perform various tasks.

Now, Internet users can use voices to perform Google searches among other services. “If there is no Kinyarwanda data set and voice is being used to search then it means people need to adapt to the language that is being used for search,” said the tech enthusiast from Digital Umuganda stressing the importance of building a Kinyarwanda voice-based system.

Other solutions that could be built through the Kinyarwanda voice system will be applied in call centres, mobile phones, transportation, applications like real time translators and voice-enabled digital assistants among others.

Developers behind Digital Umuganda need an extremely large amount of voice data to build the digital infrastructure based on a voice system, targeting 10,000 recorded voices.

According to Mozilla, this is the number of hours required to train a production speech-to-text system.

The project managers hope to complete the data set by the end of the year. This will be done through four digital Umuganda exercises.