When it comes to machine translation — literacy is key

The best way to approach machine translation — Google Translate, for example — is from an informed and critical perspective. That’s according to Lynne Bowker, the latest participant in the Concordia Library researcher-in-residence program.

Bowker is a professor at the University of Ottawa’s School of Translation and Interpretation with a cross-appointment to the School of Information Studies. She will act as researcher-in-residence from June to December 2019.

Her project while at Concordia will focus on machine translation literacy in the context of scholarly communication. As part of her residency, she will conduct research with non-anglophone student and faculty participants from Concordia to understand how they are engaging with machine translation with respect to scholarly communication and how it can be improved.

In addition, Bowker will work with Concordia librarians to develop a workshop on machine translation literacy for the university community.

Now in its third year, the researcher-in-residence program seeks to foster a strong research culture and promote evidence-based librarianship at the university.

It offers the opportunity for librarians, archivists, scholars or doctoral students to focus on an area of inquiry in a supportive and enriching environment, and to interact with Concordia Library staff and resources.

My goal is to help the scholarly community avoid the potential pitfalls of translation technology while optimizing its strengths

Can you tell us the meaning of machine translation literacy?

Lynne Bowker: On the surface, free online tools such as Google Translate seem straightforward. Open the application, copy and paste your text, choose your languages and click “Translate.”

Using machine translation (MT) technology may be easy, but using it critically requires more thought. Have you ever considered what happens to the text that you paste into the tool? Maybe you think it just disappears once you close the window? Spoiler alert: it doesn’t. Confidentiality and privacy are some issues worth thinking about before choosing to use an online MT system.

And how reliable are the translations that come out? The most recent AI-based approaches use a technique called machine learning. Essentially, the computer program is fed millions of words of text and it uses this “training data” to learn how to translate new texts.

Depending on the texts that are used for training, the machine translation system might learn inappropriate things. For instance, there are already numerous reports of systems that produce texts with gender or racial bias.

Learning how to prepare texts in a machine translation–friendly way can improve the usability of the translated output. Thinking about whether, when, why and how to use machine translation is part of what I term “machine translation literacy.”

What inspired your interest and research in this area?

LB: As an anglophone, I am able to do most of my scholarly communication in my native language. The more interactions I had with international students, visiting scholars and colleagues at international conferences, the more I began to realize my privileged position.

Then I read an account by a group of Spanish researchers who documented their struggle to have their work published in English. They reported a success rate of less than 25 per cent. The science was fine, but it just wasn’t being linguistically packaged to the liking of the English-speaking journal editors.

It would be easy to see this as a problem for non-anglophones, but I think that English speakers have a responsibility in this situation too. If we want the best and brightest minds on the planet working together to solve problems such as climate change, cancer and energy crises, then we need to make sure that researchers from all linguistic backgrounds can effectively share their research findings with one another.

My goal is to use my own expertise in the area of translation technology to help all members of the scholarly community improve their machine translation literacy. That way they can avoid the pitfalls of this technology while optimizing its strengths.

I believe this research has potential to help others beyond the scholarly community too. For instance, a machine translation literacy program could be adapted for school kids or newcomers to Canada who speak languages other than English or French. Machine translation can make a significant contribution to achieving social justice.

What are a few misconceptions about your research?

LB: One big misconception is that machine translation is “easy.” But using the technology thoughtfully requires us to act thoughtfully, rather than acting on autopilot.

Another related problem is that people think machine translation is fully “automatic.” In most cases, it’s better to treat machine translation as a tool that can help with translation, rather than as one to do translation.

Human intervention, before or after the machine translation phase, can make a huge difference to the eventual quality of the text.

Another issue that most users don’t think about is that there are human beings behind the machine.

Even AI-based machine translation systems would not be able to function without the contributions of tens of thousands of professional translators. So there’s an ethical issue to confront: professional translators are not getting any recognition for their work because all the credit is going to the machine.

Finally, English-speaking scholars don’t always realize the critical role that they need to play to begin to level the playing field in the world of scholarly communication. It’s not a problem that non-native English speakers should be grappling with alone.

What do you hope participants will gain from a machine translation literacy program?

LB: My wish is for people to come away with a better idea of how machine translation works so that they can make informed decisions about whether to use it in a particular instance. And if they do use it, I hope they will be able to do so in a way that will maximize its effectiveness and minimize potential risks.

I hope they will also give a silent nod to the professional translators and software developers who have made this type of technology possible.

What is the best way for Concordians to connect with you?

LB: I would love to hear from Concordians who have an interest in multilingual communication, translation technologies, scholarly communication or other related areas. I have an office in the Webster Library (LB-505.02). You can also reach me at 514-848-2424, ext. 7758, or by email. Email usually gets the quickest response.

Find out more about Concordia Library’s researcher-in-residence program.