Swiss AI model able to translate Romansh language
Researchers at the University of Zurich have developed an artificial intelligence (AI) model capable of translating texts from German into different Romansh idioms.
+ Get the most important news from Switzerland in your inbox
Romansh is considered particularly complex for AI.
The Alas model is the first to support Romansh idioms, the university announced.
Romansh has five regional varieties: Sursilvan, Sutsilvan, Sursmiran, Puter and Vallader. In addition to these, there is Rumantsch Grischun, which is used for drafting official texts. For AI systems, this variety poses a particular challenge.
While AI assistants such as ChatGPT already master Rumantsch Grischun to some extent, they have so far had great difficulty with regional idioms. On the initiative of the Lia Rumantscha (the umbrella body of Romansh language associations), the research team has now developed an AI model capable of handling these variants.
Results verified with native speakers
Alas was specifically applied to Romansh idioms. The data included, among other things, contributions from the Romansh radio and television station RTR, newspaper articles, literary texts, school textbooks and documents from the communes and canton Graubünden.
Join the debate:
The results were verified in a blind test with native speakers. “The comparison confirmed to us that Alas produces significantly better translations than current AI assistants,” said Jannis Vamvas from the University of Zurich.
The translations would be more content-accurate and more idiomatic. Only translations made by humans performed even better.
The translation programme Alas is available to the public free of charge on a website. The underlying AI model is also accessible as open source software for experts to test it and develope further.
More
Fact and fiction about the Swiss AI model Apertus
Translated from Italian with AI/mga
We select the most relevant news for an international audience and use automatic translation tools to translate them into English. A journalist then reviews the translation for clarity and accuracy before publication.
Providing you with automatically translated news gives us the time to write more in-depth articles. The news stories we select have been written and carefully fact-checked by an external editorial team from news agencies such as Bloomberg or Keystone.
If you have any questions about how we work, write to us at english@swissinfo.ch.
In compliance with the JTI standards
More: SWI swissinfo.ch certified by the Journalism Trust Initiative
You can find an overview of ongoing debates with our journalists here . Please join us!
If you want to start a conversation about a topic raised in this article or want to report factual errors, email us at english@swissinfo.ch.