The world is full of diversity. From races, skin colors to languages. But among them, the most endangered one is language.
According to the Summer Institute of Linguistics, or known as SIL International, there were at least 6,909 distinct languages in 2009. However, about 2,000 of those languages have fewer than 1,000 speakers. And out of that many languages, 43 percent of them are endangered.
What makes things concerning is that every 14 days, one language dies and go extinct because the knowledge is not passed on.
With the internet and technology in making the world closer than ever, digitization is not helping much. This is because more than half of all internet online contents we have on the web, is either in English or Chinese. This quickly makes indigenous, smaller languages and less popular languages in having difficulties to thrive.
Another example is in Australia. In a home of more than 300 indigenous languages, only 140 languages are still spoken, and only 18 are taught to children. What this means, there is a little cultural heritage and knowledge that is passed on to the next generation, most significantly, even to its own people.
This is where researchers want to help with some gifting hands.
Aided by technology, researchers from ARC Centre of Excellence for the Dynamics of Language, (CoEDL), created 'Opie', a social AI robot to revive and preserve endangered languages.
To do this, the researchers have been gathering materials and recordings about indigenous languages, mainly in Australia, for years. Here, they have have collected over 40,000 hours worth of material. One problem is, to make use of this huge amount of trove required the researchers to spend another 2 million hours to go through the tedious and laborious work.
This is where AI shines. This is where Opie goes to work.
Opie is a simple, low-cost robot, made of wood, some tablets, a speaker, a mobile router and a Raspberry Pi computer. But for its "brain", CoEDL has launched a partnership with Google to use its open-sourced AI platform, TensorFlow in an effort to revive these languages.
Here, the tens of thousands of hours of data can be effectively put to use. As a start, the AI mode has been made to understand six Australian indigenous languages: Bininj Kunwok, Kriol, Mangarayi, Nakkara, Pitjantjatjara, Warlpiri and Wubuy.
It has also added another five languages, which are spoken in the Asia Pacific.
The ambition of the project is to make Opie an AI language recognition that is designed to teach people to understand indigenous languages. With the ability to give lessons, Opie could easily be implemented into apps on phones and tablets. But the researchers are focusing Opie on children aged 2 to 5, which is regarded as the best age for learning languages.
"In the connected social media era, having your own language present online is extremely important to people," said Daan van Esch, a product manager at Google’s language team.