French push for AI auto-translation of European research

The European Union is set to launch a drive to translate more scholarly research into local languages using artificial intelligence technology.

The push to challenge English as science’s default language is being led by the French presidency of the EU as part of broader reforms of research assessment, but campaigners have warned that automation must not lead to neglect for the venues that develop native academic vocabularies.

The latest draft of the European Council’s plans for research assessment notes that Covid-19 had demonstrated the “necessity” of having quick access to new research and claims that the “semi-automatic translation of scholarly publications within Europe” using AI may have “major potential”.

The council says that the bloc could use the new technology to translate a “sample of the research outputs published by European researchers selected on the basis on their societal relevance”.

The French presidency “has of course a keen interest in pushing the French language, after Brexit and so on”, said one EU diplomat, who asked not to be named.

“If we want to have a genuine, living vocabulary about research and research outputs, we need to have researchers who actively work and publish in multiple languages,” said Henriikka Mustajoki, secretary general for open science at the Federation of Finnish Learned Societies and a drafter of the Helsinki Initiative, which has won the support of hundreds of organisations with its call for equal access to knowledge and protection for publishers of locally relevant research.

But she warned that while auto-translation should be part of the solution, it could go only so far. “What we’re concerned about is the sort-of technocratic solution to a very complex issue,” she said.

Software has tended to struggle with Finnish, which has distinct roots to most European languages. One solution is the Helsinki Term Bank for the Arts and Sciences, a community effort to catalogue and create a Finnish academic vocabulary.

“We have learned with this term bank that it’s not really possible to automatically extract and translate terms. We really need also the human capacity,” said Tiina Onikki-Rantajääskö, a professor at the University of Helsinki and director of the bank, explaining its dedicated disciplinary committees.

Recent bank creations include energiasääennuste: a solar and wind energy production forecast, a hot topic in a country that faces both nightless summers and sunless winters.

Professor Onikki-Rantajääskö said that politicians should think about paying the academics who work for journals such as Virittäjä, which she once edited and has published in Finnish for 125 years.

A recent shift to open access has hit the subscription revenues of some smaller journals, raising questions about the sustainability of some native language venues.

“That’s a really big problem. Even in Finland, the government doesn’t realise how important it is that it should support academic publishing in national languages,” Professor Onikki-Rantajääskö said.

ben.upton@timeshighereducation.com