Economic developers donate to speed voice biometrics project for African tongues

Bill Gates’ philanthropic foundation and two international development groups are making a modest investment in efforts to expand the languages understood by speech and voice recognition apps.

It is a growing segment of biometrics. At least one other big name is trying to speed speech AI.

The trio has pledged $3.4 million to the open-source Mozilla Common Voice project as it expands a voice dataset for a prominent East African language. Mozilla, which has compiled datasets for 60 languages, wants to create open depositories for all languages.

The funders are the Bill & Melinda Gates Foundation, the UK’s Foreign Commonwealth & Development Office and the Deutsche Gesellschaft für Internationale Zusammenarbeit.

Mozilla is the nonprofit that created the Firefox browser. The group works to see that the Internet is a public, open and accessible resource for people around the globe.

People can contribute their voices at the project site. Researchers, developers and businesses can use the resulting datasets to train products and services.

Kiswahili, the language benefiting from the funding, is spoken by the Swahili people, who live principally in Kenya, Tanzania, Uganda, Rwanda and Burundi. Mozilla says about 100 million people speak the language, but estimates vary greatly.

Mozilla claims that no native African languages are spoken by products from Amazon, Apple or Google.

Earlier this month, executives in Facebook’s AI lab said they have developed speech tools that are trained without supervision.

Not only is the advance billed as making the creation of speech recognition systems faster and less expensive, but it is also supposed to do so for all languages and dialects.

Article Topics

Africa  |  biometrics  |  dataset  |  Gates Foundation  |  Mozilla  |  open source  |  research and development  |  speech recognition  |  voice biometrics  |  voice recognition