[ad_1]
When discussing Internet connectivity in Africa, the conversation usually revolves around those who have access to it and those who do not, namely the digital divide. However, a new study indicates that there is also a linguistic gap, a gap that has widened, especially with regard to speech recognition technologies.
While voice interfaces like Amazon's Alexa, Apple's Siri and Google's wizard become ubiquitous, more and more people are using smart speakers to shop, set reminders, and get answers to simple questions. but essential like the weather. Research shows that half of all searches will be based on the voice of 2020, and this mbadive hub to voice commands should create an entire ecosystem of applications and interactions.
Still, research conducted by the digital financial services lab and the Caribou Digital consulting firm show that developers are focusing their efforts on improving language skills in English and less on developing country languages. Africa and Asia. African languages are already at a disadvantage online with huge platforms, including Twitter and Google AdSense, which do not support any African language.
And while companies have made enormous strides in automatic natural language processing (NLP) or the ability of computers to understand human voice and language, a gap is emerging that puts low-income and low-income populations in the background. widely spoken. As a result, this disparity will not only create a gap between those who can use this form of artificial intelligence for communications and those who can not, but will also impede the ability of users to exploit these applications for development interventions such as health and finance.
One factor that influences this biased attention is profit, with language-conscious NLP providers who could help them make money. By multiplying the number of speakers of one language by the gross domestic product per capita, the authors found that the 100 most spoken languages accounted for about 96% of world GDP. Yet these 100 languages represent less than 60% of all populations, highlighting "a fundamental tension between the commercial and social value of languages".
African languages also lack data to form machine learning systems, discouraging researchers from starting from scratch. Although languages like Swahili are spoken by nearly 100 million people, there is no well-known repository that can be used to power speech recognition software. The study notes that the creation of accurate machine translation services requires about 100,000 hours of recorded speech.
And while the rapid growth of email applications has increased the amount of content available to researchers, it is largely generated in so-called "obscure social" applications such as WhatsApp and Facebook Messenger, where shared information can not be measured or measured. nor used as a set of data to create sophisticated systems.
The multilingual nature of modern Africans also means that developers do not deal with mixed linguistic models. The code change is inherent to many countries, users speaking or writing in two or three languages in a publication or conversation. An example is the combination of English and Swahili in Kenyan sheng, the mixture of French with Arabic or Berber in Algeria, or the Romanization of Amharic.
However, these shortcomings have not prevented start-ups from innovating voice services for African customers. Farm.ink has developed chatbots that allow Kenyan farmers to receive information based on user-generated messages. The financial technology company Teller also integrates financial services to the courier of customers from Madagascar and West Africa.
These examples are relatively nascent and are always oriented to English and French, respectively. But they offer an illustration of how voice technology could fundamentally change the way Africans access the Internet.
Sign up for the Quartz Africa weekly report here for news and badysis on African business, technology and innovation in your inbox
Source link
