September 6, 2022
A recent article on ScienceDaily discussed how machine translation could make English-only science accessible to all researchers and readers around the world.
Students and scientists at UC Berkeley, have said of the currently available machine translation systems that “though flawed, they have become good enough for researchers to broadly translate their work into other languages”.
This is a most welcome observation that shows proof of the increased quality of machine translation in recent years.
Language has always been a major challenge for scientists working in a language other than English. On one hand, most available research is done in English, requiring an understanding of the language. On the other hand, there is also pressure to have one’s own research made available in English in order for people to have access to it.
Translation is thus a necessity for scientists looking to keep up with research discourse in a global setting. But as we all know by now, while human translation is the gold standard, it can be an expensive undertaking, limiting the amount of research that could be translated.
In the past, machine translation was unthinkable as an alternative. The quality simply wasn’t there yet. But something changed around 2016—the introduction of neural machine translation.
Neural machine translation is a technology that uses highly advanced machine learning models based on networks that mimic the function of the human brain. In less than a decade, it has increased the quality of machine-translated output by leaps and bounds.
Learn more here: Everything You Need To Know About Neural Machine Translation
Suddenly, people were beginning to find machine translation useful for a wide variety of use cases, and translating research was one of them.
But this isn’t to say that machine translation is some magic pill—naturally, the technology still has its flaws in terms of accuracy. But it’s a far cry from where it was before.
When we talk about machine translation, the first thing that comes to mind is usually Google Translate. There are also other tools like DeepL, Amazon Translate, and many others today. But what people don’t often realize is that even among these options, there’s a wide range of customizability.
Most major machine translation engines come with a free version that can be accessed by anyone. These versions are trained on generic language data, which makes them best suited for general-purpose translations.
Scientific research is full of jargon and domain-specific terms that the stock models of machine translation engines are unable to handle well. Customized versions can incorporate glossaries and translation memories to handle these terms and allow for even more accurate translations for scientific research.
Thanks to advances in machine translation technology, more and more scientists are able to make their work available in languages other than English. They are also able to access a wider range of knowledge resources in their domain.
But there’s also another, less obvious way machine translation can make science more accessible.
Scientific research can often seem intimidating to the average reader (and perhaps even scientists) because of the jargon and often complex language that they are written in. There’s always a call to write in simpler language that can be understood by a wider range of people.
With machine translation in the picture, researchers and scientists now have additional incentive to do so, as simpler language also works to the advantage of machine translation systems. The clearer the language, the better the machine translated output, which is a win all around.
As it stands, not everyone is aware of the current possibilities regarding the quality and usability of machine translation, but this is beginning to change. And machine translation is improving at an accelerating pace.
This is something that we believe stakeholders in scientific research need to take into more serious consideration in the way they operate. Research departments and institutes should lead the way in investing in machine translation resources, as translation creates a healthier research community, and machine translation has great potential to contribute to this endeavor.