Mozilla Swahili

Mozilla Common Voice Picks Up $3.4M to Teach Voice Assistants to Speak Kiswahili

A coalition of international non-profits is investing $3.4 million in Mozilla Common Voice to lay a foundation for voice assistants that speak Kiswahili. The open-source database of cataloged and transcribed voice recordings will apply the money toward building a robust voice dataset in the East African language, known elsewhere as Swahili, which about 100 million people speak, but has yet to be addressed by the major voice assistant developers.

Uncommonly Common Access

Amazon Alexa, Google Assistant, Siri, and other voice assistants have steadily added new languages to their repertoire since their introduction, no African language has ever been included in their plans. The geographic range of Kiswahili makes it an excellent candidate for Common Voice. Kiswahili is widely spoken across Kenya, Uganda, Tanzania, Rwanda, Burundi, and South Sudan, with pockets and of native speakers all over the world. To teach voice assistants to converse in a language requires huge datasets of people speaking in that tongue. The tech giants usually hold that data tightly, leaving others to look for other sources, which is where Mozilla comes into play.

Mozilla Common Voice began in 2017 as exactly that kind of alternative to major corporations for those looking for the data needed to train a voice AI. Common Voice is also a source for making voice assistants more accessible to those whom the existing voice assistants may have trouble understanding. The funding will make Kiswahili a significant part of the project. The investment comes from the Bill & Melinda Gates Foundation, the Deutsche Gesellschaft für Internationale Zusammenarbeit (GIZ) GmbH (German Development Cooperation), and the UK’s Foreign Commonwealth & Development Office (FCDO). The cash will help with hiring a handful of people to focus specifically on the Kiswahili database and with the outreach and technical assistance needed to streamline getting volunteers to contribute their voice samples to the database.

“Language is a powerful part of who we are, and people, not profit-making companies, are the right guardians of how language appears in our digital lives,” Mozilla Foundation Special Advisor for Africa Innovation Mradi Chenai Chair said in a statement. “By making it easy to donate voice data in Kiswahili, Common Voice will support East Africans to play a direct role in creating technology that helps rather than harms their communities. We are thrilled to join with partners who share Mozilla’s vision for helping more people in more places to access voice technology.”

Common Talk

The Common Voice Database is enormous already, with more than 9,000 hours of 60 different languages. It’s the largest public domain voice dataset in the world. Common Voice has also proven it can grow quickly, as it was at 7,226 hours in 54 languages a year ago when Mozilla released a massive data set of voices,. If the Kiswahili project goes well, Mozilla may start applying a similar approach to other languages currently unrepresented by the major voice assistants. Common Voice is partnering with African companies and research groups to leverage the database into voice tech for the area, especially financial and agricultural services.

The funding and Kiswahili focus come a little over a month after Nvidia made a $1.5 million investment in Mozilla Common Voice and started working with Mozilla on voice AI and speech recognition. That investment also marked Common Voice’s transition to be fully under the Mozilla Foundation umbrella.

  

Nvidia Invests $1.5M in Mozilla Common Voice Open-Source Project

Mozilla Updates Massive Open Source Voice Data Collection

Mozilla Officially Shuts Down Firefox Voice Browser Extension