One of the newly included languages is Northern Sami. – I think it is unique and exciting. Development is happening so fast here that I think we will see very good results in the small languages, says technology enthusiast and former Sámi Parliament director Rune Fjellheim. “Shit in, shit out” On the other hand, the Sami language technology community at the University of Tromsø (UiT) is not as keen on Google’s new grant. Google’s language tools are made with artificial intelligence (AI). To create such a “translation machine”, a lot of data is needed, and traditionally an enormous amount of parallel texts has been needed. Technical breakthroughs have meant that less data than before is needed to reach Google’s quality standard when it comes to machine translations. Inga Lill Sigga Mikkelsen says that the Sami language community at UiT has deliberately avoided creating machine translators into Sami. Photo: Vetle Knutsen / news Inga Lill Sigga Mikkelsen, linguist and senior engineer at Divvun, nevertheless believes that the machine translation is too bad. She fears that KI will produce Northern Sami texts with bad language, and that people will publish them online. They are worried that these “bad” texts will again be used to train the Sami machine translators, and that this could “contaminate” the data. – If you feed bad language into AI, then it becomes “shit in, shit out”, says Mikkelsen. Precisely for this reason, the Sami language technology community at UiT has deliberately avoided creating programs that translate into Sami. – Good news for small languages Sondre Ronander, head of communications at Google Norway and Finland, disagrees with the language community at UiT that there is not enough data to create good translation programs. – It is precisely because of the alternative method we have used here that we do not need the same large amount of parallel texts as previous methods have required, he says. Communications manager at Google believes the new addition is good news. Photo: Mari Reisjå / news Ronander says that Google has taken into account the language’s vulnerability, and points out that they have high quality requirements before launching translators into new languages and that people with Northern Sami language skills have been involved. – I think this is great news, good news for small languages. I hope it can help prevent the linguistic heritage of many more languages from disappearing, he says. Nevertheless, he points out that translation programs will never be 100 percent correct. – I don’t think you should ever trust 100 percent of any computer system. We must always have source criticism, also in digital life, says Ronander. What do you think of Google Translate in Northern Sami? It’s awesome! I don’t like it. Do not know. Show result Thinks it is wrong to demand the perfect Rune Fjellheim is not afraid that machine translators will damage the language. – I think that the most important thing is that we actively maintain the language in interpersonal relationships. And then this becomes a kind of aid, he says. In addition, he believes that technology is difficult to stop, and that it should be embraced, rather than feared. – I just want to encourage people to use these tools. – And perhaps assist in correcting the translations that you see are obviously wrong. Because it helps the next user, he adds. Rune Fjellheim is a former director of the Sami Parliament, and calls himself a super user of AI and technology, particularly within Sami applications. Photo: Nelly Anna Karolina Engström / news He believes that you cannot wait to use language tools until they are perfect. – Here I think you make a big mistake if you demand perfection before you get started, says Fjellheim. He compares it to learning a new language. – You speak a new language a little poorly, before you get better. This system just keeps getting better and better. Especially if we can contribute a little now and then by correcting a mistranslation, says Fjellheim. Strong appeal Inga Lill Sigga Mikkelsen has a strong appeal to use AI translations with caution. – Do not translate into Sami with Google Translate and publish it without quality assurance. It’s going to be really scary. – This is what will be the training data of the future for KI. We want good AI training data, she says. If people follow this, she believes that the translator can be positive. – I think it is positive for Sami society, but also for Norwegian and people in the world, if you don’t use it to contaminate what is published. – But we don’t know how people will use Google Translate. Time will tell if it goes well or not. Why has the Sami language community not made a translation into Sami? The Sami language technology community at UiT has deliberately avoided creating programs that translate into Sami. They believe that the quality is not good enough, and they fear that people will use machine translators without proofreading texts before they are published. – The languages are too far apart. There will be lots of mistakes. We think it might do more harm than good, says Inga Lill Sigga Mikkelsen. Mikkelsen believes that a machine translation must be very good for Sami speakers to choose to read machine-translated texts in Northern Sami, instead of the original in Norwegian. The Center for Sami Language Technology at UiT, on the other hand, has created a tool that translates Northern Sami texts into Norwegian, so that those who can write in Sami can use the language to a greater extent. – Isn’t it better if the Sami language technology communities create a translator into Sami themselves, in order to ensure their quality more? – It is a point that they are now out there, says Mikkelsen. – If others start doing it, it might force us to do it, she adds. Published 07/07/2024, at 08.32 Updated 08.07.2024, at 11.50
ttn-69
Google has launched a translator from Norwegian to Sami – news Sápmi
