Language is a major barrier to internet accessibility in Africa

Increased internet-connectivity might not be translating into increased internet-accessibility, primarily due to the fact that internet-based tools such as voice recognition are not available in most African languages – according to collaborative analysis from consultancy firm Caribou Digital and the Digital Financial Services Lab.

Much has been made of the rapid spread of internet connectivity across Africa. Some studies predict that Africa will have as many as 1 billion internet connections by 2022, which is expected to boost the continent’s digital ecosystem in general. One reason for this rapid spread of internet access has been a sharp drop in the cost of internet connections. 

The low cost of internet has created a more equitable environment on the one hand, as individuals from all socio-economic backgrounds can now make the most of digital applications. However, new analysis shows that there remain barriers to a truly equitable internet environment.

Farnham–based investment advisory and research firm Carabou Digital has conducted a comprehensive analysis in collaboration with the Digital Financial Services Lab with a focus on the language barrier that emerges when most internet applications are developed only in a handful of languages.

Voice recognition technologies such as Alexa by Amazon and Siri by Apple are rapidly gaining popularity, so much so that 50% of all search functions will be voice-based by as early as 2020. Much like most other internet applications, however, these tools neglect most African languages.

As of this year, there are over 98 million Swahili Speakers in Africa and over 63 million Hausa Speakers. Other languages with tens of millions of speakers on the continent include Yoruba, Oromo, Zulu, Igbo, Amharic and Berber, most of which are absent from internet applications.

According to the firm, this situation appears to be far from rectification, given that most companies that are developing internet-based tools continue to focus on perfecting their capabilities in English. One reason for this gap in priorities is that English-speakers in Africa constitute the high-income population, which is the area of most potential profit.

Nevertheless, there are also more practical barriers to developing such technology to function with other African languages, including the fact that there is a lack of data available to develop machine learning in more remote languages. The development of these technologies requires approximately 100,000 hours of recorded speech, which is lacking for most African languages.