The Linguist

14 The Linguist Vol/61 No/2 2022 thelinguist.uberflip.com ONLINE INCLUSION Kọlá Túbọsún considers technology issues in African languages and why the ability to write online is crucial O ne of the first events that alerted me to a limitation of African languages in technology was a small wriggly line. I was in the first year of university as a student of linguistics, and I had obtained my first computer. The wriggly line was red, as will be familiar to users of Microsoft Word, and marked a word that was not in the dictionary. In this case, however, the words were my name, some places in Ìbàdàn, where I lived, and occasional words of Yorùbá that I needed to use in the essays I was writing. This was in the early 2000s. Many years have passed since then, and options have been created to add words to the dictionary of whatever program you are using so that they are recognised. But I quickly realised that this was not the only obstacle that I had in using the computer as a bilingual speaker of English and a Nigerian language. Yorùbá is a tone language that needs diacritics on and under vowels in order to disambiguate meaning or indicate vowel quality. In a notorious example of tonal ambiguity in Yorùbá, one word can mean 'penis' (okó) or 'farm' (oko) or 'stone' (òkò) or 'husband' (ọkọ) or 'vehicle' (ọkọ̀ ) or 'spear' (ọ̀ kọ̀ ), depending on how it is written. There was no word processor on any of the computers I had access to that could write with the appropriate marks. Several books published in the language had to be hand-marked after they had been printed. Because tone is an integral part of the language, if we want it to survive in the electronic space, it is important that these diacritics can be written on a computer. And so in 2015, when I launched the YorubaName.com dictionary of Yorùbá names, as the first publicly accessible, free and crowdsourced project of its kind, we also created tone-marking software. Available for free download, it allows people to write Yorùbá on any platform. It has now been downloaded more than a thousand times. When I got to work at Google a few years later, I was involved in creating the same functionality for mobile phones through the 2017- released GBoard App, which benefits thousands of languages around the world. The need for reliable sources There is a second reason why it is vital for African languages to find a place on the internet, beyond the most important one of representation. Language engineers typically use available text corpora from existing websites when they want to create language tools, assuming that these texts best represent how the language is written. As the English language makes up a huge percentage of the texts online, one can be sure that much of the text one finds in English is a representative sample of the language, needing very little correction before use. With Yorùbá, Igbo, Hausa and many other African languages this is not the case. Until BBC Yorùbá started publishing on the web in Nigeria around 2016, there was no reliable place online to find a properly written Yorùbá text. This certainly contributed to the absence of a lot of speech products in the language. Despite some positive changes, even today there is no ATM in Nigeria that can be used in a local language. No phone that can be fully operated by voice in a Nigerian language. No voice-enabled electronic device that can be controlled in any African language, as far as I know. This has significant implications for financial inclusion, technological inclusion, literacy and societal equity. I met a visually-impaired Nigerian recently who can use his iPhone by utilising the accessibility options, but if he couldn't speak a word of English – like many people on the continent – this wouldn't be possible and he would be severely disadvantaged. In 2015, I led a team creating a Nigerian English accent for Google Maps and Google Assistant, giving many more Nigerians access to a voice that sounds like Africa represented ACCESS DENIED Missing diacritics change the meaning of Yorùbá words, which can render the text unintelligible ` ´

• Cover