Machine translation could make English-only science accessible to all
Machine learning using artificial intelligence has improved computer translation over the past decade, but scientific articles employing specialized jargon are still a challenge for machine translation. Nevertheless, scientists should prioritize translating articles into multiple languages to provide an equitable landscape for budding scientists worldwide, UC Berkeley researchers argue. Credit: Valeria Ramírez-Castañeda, UC Berkeley

The first English draft of scientific papers was generated by Chinese researchers in a lab at Beijing's Beihang University, and it was surprising to see that they used a translation service. Most of the high-profile journals are written in English.

Postdoc's used to translate everything and then modify and polish it using a translation service. The paper didn't make sense after the first translation, according to a rising junior at the University of California, Berkeley. All the words were stuck together in random order.

She believed that there had to be a better way.

She signed up for the new seminar when she saw it last year.

Tarvin's trial balloon was a class that will be taught at UC Berkeley for a third time. With renewed campus-wide interest in diversity, equity and inclusion, she and working groups within her department thought that the class could help UC Berkeley address a long-standing issue in science: English is a major obstacle to scientists who are not native English speakers.

Foreign students and scientists are not the only ones who are at a disadvantage when it comes to communicating science in English. Many American students are also American born. Within the University of California system, 39% of first- generation students grew up with a language other than English as their first language.

Tarvin said that many of their students were born in California. Since they were young, translation has been a part of their lives.

Tarvin was able to teach students skills in translation literacy and encourage them to be activists in the realm of structural change by taking the class. The students seem to agree that addressing language barriers is important after taking the course and I have seen a positive reception of this sort of activism from them.

The class led Tarvin and some graduate students at UC Berkeley to write a scientific paper evaluating new machine translation tools that can be used by people worldwide to make their scientific articles accessible to non-English speakers. The paper was published in the journal. There are online translations of the languages of the co-authors.

Tarvin said that the goal is to give people the tools and motivation to translate their own scientific research. Science can be based on more than one language. There are a lot of benefits that come from using multilingual approaches. Better science communication will benefit society by publishing in more than one language.

The first author of the paper is a UC Berkeley graduate student in environmental science, policy and management. It is a barrier that we can overcome with this new technology. We explain about the technology and how it can be implemented and the things that we need to be aware of when we use the technology, as well as all the wonderful and positive ways that science communication can be transformed by bringing this new technology to bear.

The network should be multilingual.

Computer translation used to be a joke. People shared examples of mistranslations that seemed to mock other cultures.

Machine learning has made it possible for tourists to communicate with people in other countries on the internet.

For text that contains a lot of jargon, it's not a good idea to use a translation service.

One of the graduate student mentors for the course said that the translation quality is not for a journal. It's very expensive to get a professional translator for your work.

There are a lot of free services that can convert English scientific writing into other languages. DeepL, which uses neural networks and claims to be many times more accurate than competitors whenTranslating English into Chinese, Japanese, Romance languages or German, is one of the services included in these.

Translation is reaching more and more people. Whether or not you are an expert, bilingual, or both, the ability to translate is so fast because of the technologies we have today. How can we integrate this into our work as a scientist, and how do we change the expectations for scientific communication?

Machine translation could make English-only science accessible to all
El aprendizaje automatizado que usa tecnologías de inteligencia artificial ha mejorado la traducción en computador en la última década. Sin embargo, los artículos científicos que emplean terminología especializada siguen siendo un reto para la traducción automática. No obstante, la comunidad científica debería dar prioridad a la traducción de artículos en varios idiomas para ofrecer un panorama equitativo a los científicos y las científicas en formación de todo el mundo, afirman los investigadores de la UC Berkeley. Credit: Valeria Ramírez-Castañeda, UC Berkeley

The English language is used in science.

Tarvin's interest in translation came from one of her graduate students, Valeria Ramrez Castaeda, who in 2020 published a paper describing the costs incurred by her fellow Colombian PhD students who wanted to publish or interact with colleagues in a world dominated by English.

Tarvin decided to focus her new seminar on translation of papers in the fields of evolution and ecology because she was interested in how some animals used poison. She looked for students and mentors who were bilingual or multilingual.

Everyone in the class has a family relationship with language.

Mairi- Louise McLaughlin, UC Berkeley professor of French and linguistics and an expert on journalistic and literary translation, was asked by Tarvin to speak to the class about how professionals approach translation. When the students tried to translate scientific papers, they found that it was related to that topic.

The rising sophomore took the course in the spring of 2022. When rendered into Chinese, English's long, complex sentences and use of multiple words didn't make sense.

She said that they don't do that in Chinese because it's very tedious.

Many English translations get garbled without much polishing.

People will get annoyed if you write a confusing paper because they can't understand what you're doing. That will have a huge impact on how people evaluate the research or not. That's a big problem in the scientific world.

Writing scientific papers in plainer English benefits both English and non- English speakers.

"If your first language is not English, and you're just trying to read the English language version of the paper, it will feel less ambiguous and more readable when the writer has used plain language." When you translate something written in plain language, the machine learning tools will be able to do a better job. If someone wants to translate your writing into a million languages, they'll have a much easier time with it if it's written in that way.

There are obstacles to widespread translation of scientific articles, including where to find them and how to deal with copyrighted works. Most journals don't accept articles that aren't in English and don't allow translation of articles. Tarvin found that few journals have any policies about translations and that many publishers charge high fees to post a translation online after publication.

Tarvin said that many journals don't allow you to freely publish translations after publication and that few have platform support where you could have an abstract in a second or third language. The web platforms are a major barrier to this, not just the publishing and copyright rules.

The more translations there are, the more material there is for training machine translation systems to do better jobs.

She said that people in Emma's lab are doing the same thing she is doing in her lab. Sharing our positive attitude towards this and how it can make a difference for people has influenced a small group of people who are starting to incorporate translation into their scientific work.

Valeria Ramrez-Castaeda and Andrs Bldi are two of the additional co-authors of the paper.

More information: Emma Steigerwald et al, Overcoming Language Barriers in Academia: Machine Translation Tools and a Vision for a Multilingual Future, BioScience (2022). DOI: 10.1093/biosci/biac062 Journal information: BioScience