One year ago this month, Google Translate added Kinyarwanda — a language principally spoken in Rwanda, Burundi, the Democratic Republic of the Congo, Uganda and Tanzania. And despite a year to perfect it, Kinyarwanda speakers say the tool’s translations are still “hilarious,” reports news site The East African.
According to The East African, Google Translate has problems with basic phrases such as “nagushatse ndakuburaa” — which means “I looked for you and I couldn’t find you.” The tool’s English translation is “I married you, I missed you.”
Slang is also problematic for the free, online machine translation (MT) system — as it also is for human translators working in many languages. Translating Kinyarwanda slang can be even more difficult though, explains freelance translator Jean Bosco Iradukunda, as the language isn’t represented that much online. According to Web Technology Surveys, fewer than .01% of content on the world wide web is available in Kinyarwanda. MT engines rely on both previously translated and natively in-language materials to learn. If information in a certain language is limited, its MT production output and quality will be as well. With the case of Kinyarwanda specifically, The East African speculates the issue stems from a shortage of online, original language content.
“Kinyarwanda lacks wider usage on the internet, making it disadvantaged. We have to digitize our culture, publish in Kinyarwanda, poetry, folklore, blogs,” Bosco Iradukunda told The East African.
Four other languages of limited diffusion (LLD) — also known as minority or rare languages — hit their one year Google Translate anniversaries in February: Odia — spoken in the Indian state of Odisha also known as Oriya; Tatar — spoken in west-central Russia, Romania, Bulgaria, Turkey and China; Turkmen — spoken in Turkmenistan, Kazakhstan and Uzbekistan; and Uyghur — also known as Uighur and primarily spoken in the Uygur Autonomous Region of Xinjiang, located within China.