Exclusion machine?
Voice recognition to the challenge of regional languages
Voice recognition, a technology with strong prospects for development.
Getty Images - LeoPatrizi
Text by: Dominique Desaunay Follow
3 min
Online voice search is a voice communication device that allows you to give orders to a tablet, smartphone or virtual personal assistant.
This technology is gaining ground around the world, but some language communities fear being excluded from it.
Publicity
Read more
While the web may seem very inclusive, it might not be that much.
Currently, just under 50% of the world's population still live without the possibility of an internet connection.
Moreover, emerging economies and marginalized linguistic communities are often the last to gain access.
This is the reason why the Mozilla Foundation launched
Common Voice
, a
crowdfunding
project on a dedicated site to create a free database allowing the development of speech recognition software in order to defend a web that would be accessible and open to all.
Feed the system with voice data
Concretely, Common Voice has launched an initiative which invites Internet users to donate their voice data.
The objective of this operation is to develop various automatic speech transcription services thus offering an alternative to Siri, Alexa, Google Home, which, like many other high-tech conversational devices, are managed exclusively by large American firms. .
"
We plan to collect up to 10,000 hours of voice in all existing languages,
" says the foundation on its site, which advises Internet users to offer a variety of voice data by speaking as naturally as possible without trying to hide an accent or them. familiar intonations specific to our conversations.
To carry out its project, the foundation needs data from the voices of women, men, children and the elderly, in order to gather all the diversity and richness of a language.
The fact of having several tones of voice will make it possible to humanize the vocal devices, estimates Mozilla which recalls that this vocal data is entirely "
anonymized
" before making them available to young shoots, linguistic research centers, and even to its own computer engineers responsible for developing future voice recognition applications on their internet browser.
A voice to break the isolation
To fully understand the challenge of this technology, it should be noted that in these times of pandemic, the voice is on the rise compared to the written word or the tactile, which has hitherto been favored in digital uses.
Podcasts, vocalization of instant messages, chatbots and social networks to chat with others have helped to break a little the isolation imposed by repeated confinements.
Ultimately, a whole new electronic communication based primarily on audio is emerging.
On the condition, however, that voice recognition systems finally understand the subtleties of your mother tongue regardless of its origin, accents or even the rarity of its use.
Newsletter
Receive all international news directly in your mailbox
I subscribe
Follow all the international news by downloading the RFI application
google-play-badge_FR
Internet
Technologies
Social networks