Google Translate adds Inuktut language to its online service
Latest addition marks first Indigenous language in Canada to be added to the platform
Inuktut and English are regularly seen together throughout Nunavut and Nunavik. But Google Translate is making it easier to translate between Inuktut and other languages by making it the first Indigenous language in Canada to be added to the online giant’s translation service. (File photo)
The Inuktut language, spoken by approximately 40,000 Inuit in Canada, is now available on Google Translate, a service used to translate text, images, documents and webpages from one language into another, the internet giant announced Thursday.
This marks the first Indigenous language in Canada to be added to the platform.
Inuktut is a broad term encompassing different dialects of Inuit language spoken across Canada, Greenland and Alaska.
Isaac Caswell, a senior software engineer with Google, says the project began as part of an initiative to create an artificial intelligence language model to support the 1,000 most-spoken languages from around the world.
“The project as a whole started in 2020 as part of the next 1,000 languages initiative, that was myself and few others who felt like we could do much better,” Caswell said in an interview Thursday.
“At that time, Google Translate didn’t have any Indigenous languages from anywhere in the Americas.”
The service will give users the ability to translate Inuktut into English or English to Inuktut.
For Google, the first step in the process was training its model to recognize Inuktut.
The company reached out to Inuktut speakers, community members, and Inuit leaders for consultations, including Inuit Tapiriit Kanatami, Canada’s national Inuit organization.
ITK helped ensure the model outputted accurate translations, and also informed Google of another writing system other than syllabics for Inuktut.
“Of course, the model is going to have some errors, there’s no perfect machine translation. But ITK helped us determine whether the level of mistakes were too high or not,” Caswell said.
Inuktut uses two writing systems, qaniujaaqpait (syllabics) and qaliujaaqpait, a unified orthography that is representative of sounds across all dialects and uses the Roman alphabet.
Google Translate has both writing systems available for use on the service.
“We wanted to make sure we could fully support both scripts before releasing it to the public and not undermine any of the work they’ve been doing.”
ITK could not be reached for comment Thursday, but wrote in a statement on social media it hopes the new tool will help promote increased knowledge of Inuktut and Inuit.
Caswell said Google has looked into adding other Indigenous languages, like Ojibway or Cree, but that is currently not technically feasible.
“The amount of data, even with some translations from professional translators, was simply not enough to have a functional model,” he said.
The Inuktut option on Google Translate should be available everywhere by the end of the day on Thursday, Caswell added.
“I think it’s important for us to use our responsibility to elevate not just the groups that have already been privileged for many centuries, but also the groups that have been historically marginalized like Indigenous peoples in the Americas,” Caswell said.
It’s great news! I do hope that Canadian Inuit can agree on standardized rules and vocabulary for uses such as this one though. Because if they don’t, then these tools will decide which dialect to use and will influence the evolution of Inuktut. For instance, if you look up thank you, caribou, and other words that have regional differences, you will see that the translation is sometimes from Northern Baffin, Southern Baffin, Nunavik, Western Arctic, etc. But perhaps it’s not such a big deal, only Inuit can tell. At least having access to these options is a very good step!
Indeed.
However, it is interesting to notice that not even the name of the language is standardised. The article refers to “Inuktut”, whereas Google Translate gives the option of “Inuktitut” (as does Bing on theirs) .
So close, yet so far.
Inuktut is the umbrella term for an Inuit language. Inuktitut is a language within Inuktut. So Bing and Google are correct that the language is Inuktitut, and Nunatsiaq is also correct that Google added an Inuktut language to its service.
It’s like if there was previously no English on Google, and there was an article that said Google added English to its service, and when you went to Google it said “English (American)”. Doesn’t mean that the name of the language isn’t standardized.
Just goes to show that you’re new here.
Inuktut is just a word that means ‘any of the various Inuit languages of Canada’. It is clunky and not terribly useful, but it is what we have.
Kinship is important in family life, so I checked out some kinship terms in Google Translate. (Please excuse me if I make mistakes here, and correct me, also I apologize that I’ve never got my stubborn laptop to key up syllabics, so I have to use Roman orthography.)
I checked out two basic terms:
*brother — and Google gave both anik (used by a female for her brother) and also angajuq (used by a male for his older brother), but it did not give nuqaq (used by a male for his younger brother)
*sister — and Google gave only nuqaq (used by an older female for her younger sister) and not angajuq (used by a younger sister for her older sister) and not najaq (the term used by a male for his sister).
It’s not all bad, these are, I think, OK:
*father’s brother — aqaq
*mother’s brother — angaq
*only daughter — panituanga
*only son — irngnituanga
It looks like they need to hire an elder to correct the blind spots resulting from input by English-only staff.
There is also Microsoft translate as well through Bing. I used it yesterday for a quick translation