A language is considered endangered when it has a small number of speakers and isn’t being transmitted to younger generations. According to current estimates, a little more than 40% of the world’s approximately 7,000 languages are endangered.
If these languages cease to be spoken, there would be massive losses of cultural heritage and knowledge.
Many endangered languages either don’t have enough written material available online, or else aren’t written languages at all, making them impossible (or at least more challenging than usual) for AI to learn and translate. The situation is similar for human translators, as well. The more obscure a language, the less demand for translations, and also the less material to learn it.
The lack of available human and machine translators for endangered languages has a number of implications. One of the most troubling is what happens when important information needs to be communicated in a health crisis or the aftermath of a natural disaster.
This issue was recently in the spotlight when egregious translation errors were found in Alaskan disaster relief information distributed in the endangered Yup’ik and Inupiaq languages. Yup’ik is spoken by an estimated 10,000 people, and Inupiaq is spoken by about 3,000.
In late September 2022, Typhoon Merbok damaged the properties of more than 20,000 people living along the Bering Strait, leaving significant damage to many of their homes. Alaska’s local and federal government officials made it a point to distribute relief information not only in English, but also in indigenous languages spoken in the region, including Yup’ik and Inupiaq. But when native speakers of the latter two languages read these pamphlets, they were left scratching their heads.
The Independent’s Mark Thiessen reports that the texts were peppered with phrases that translate to nonsensical or out-of-context phrases like “Your husband is a polar bear, skinny.” One of the strange sentences randomly had the word “Alaska” inserted into the middle of it.
When experts looked into the matter, they discovered that these phrases were directly pulled from the book Yupik Eskimo Texts from the 1940’s. Why was this the case, and why were these phrases, which have no relevance to typhoons or disaster relief, thought to be useful here in the first place?
Intriguingly, no one has come forward to share exactly what went wrong. And interestingly, none of the other translations of the text had errors like this. The language service provider, Accent on Languages, has apologized and refunded their fee, but given no explanation.
So, what could have caused these translation errors?
One guess might be that the company’s translators had few resources for these endangered languages and relied on the easiest - or possibly only - one to come by.
Another guess is that none of the translators involved knew Yup’ik and Inupiaq at all, and relied on machine translation. It’s possible that AI had one print resource for the Yup’ik language, and that was Yupik Eskimo Texts from the 1940’s. Translators often use machines in a number of ways. But translators aren’t supposed to blindly trust AI and leave translations unchecked. The text should have been read over by a human being.
That said, since this text was needed quickly, in response to a crisis, maybe the translation company didn’t think it was worth taking the time to check?
Through all of this negligence runs an undercurrent of, at best, a lack of appreciation for the Yup’ik and Inupiaq cultures, and at worst, possible downright racism.
One good bit of news from the situation is that everyone was still able to get help. As Thiessen explains, most Yup’ik and Inupiaq speakers are also at least limited proficiency English speakers, and were able to use the English version of the text. Still, that shouldn’t have had to happen, especially in a country like the US, which has no official language.
There are a lot of lessons that can be learned from this mistranslation incident. For one thing, it’s a powerful reminder that no one should totally trust machine translation and not check it for inaccuracies. It also shows that when it comes to endangered languages, it’s not just the cultures they’re tied to that are at risk; in some cases, it can also be its remaining speakers’ lives and well-being.
A number of organizations, including UNESCO, are working to encourage the preservation and transmission of the world’s endangered languages. It’s a major goal that may even seem impossible. But this incident shows one of the many reasons it’s important to try. Language perservation isn’t just about our culture, but also about access to information, which should be a basic human right.