In this week’s SlatorPod, we are joined by Lynne Bowker, Professor at the University of Ottawa to discuss the Machine Translation Literacy Project, whose eye-catching infographics have proven very popular on Twitter.
Lynne shares her background working in higher education and how this led to her running the MT Literacy Project. She talks about the evolution of machine translation (MT) systems and how this inspired her to teach users in the wider population about the dos and don’ts of using MT output.
Lynne talks about the different use cases of free online MT systems, highlighting her research on how international students use MT as an aid for scholarly writing in English. She mentions the importance of confidentiality and privacy when using programs such as DeepL and Google Translate.
The pod rounds off with Lynne’s plans to train teachers in MT Literacy and introduce the project to thousands of secondary school students as part of enhancing their digital literacy skills.
First up, Florian and Esther discuss the language industry news of the week, with TransPerfect becoming the first language service provider (LSP) to report annual revenues of over USD 1bn. Another company to clear the billionaire-dollar mark was Iyuno-SDI, as the media localization provider was valued at USD 1.2bn in its latest round of investment from IMM Investments.
Over in Europe, the “Irish derogation” has ended — meaning that every piece of EU legislation must be published in Irish, giving it the same status as the other 23 official EU languages. Florian talks about a mistranslation by the Scottish Government that saw “Happy Burns Night” (a celebration of Scottish poet Robert Burns) being mistaken for “heat burns” in Scottish Gaelic.
Meanwhile, Esther covers Straker Translations’ Q3 FY22 financial update, with the LSP recording a 99% increase from the Q3 FY21 and a 26% increase from Q2 FY22.
Florian: Tell us a bit more about the University of Ottawa’s School of Translation and Interpretation. Give us a sense of the university size, courses size, faculty courses, et cetera.
Lynne: I work at the University of Ottawa, which is a bilingual university. It is the only fully bilingual university in Canada. It is in the capital city of Ottawa, so that makes sense because Canada has two official languages, French and English. We have about 45,000 students, roughly 40 or 38,000 undergrad and 6 or 7,000 grad students, so it is a decent size university by Canadian standards. Within the university, I work at the School of Translation and Interpretation. That is my home department and I have a cross-appointment to the School of Information Studies. At the School of Translation, we have programs at all three levels, so we have a professionally oriented bachelor’s program where we are intending to train students to go out and work in the professional translation world and we also have a professionally oriented master’s program that is focused on conference interpreting and then we have some research programs as well. Both the master’s and PhD program in Translation Studies are where students are entering the world of research. They do have some coursework to do, but a big focus of their program at that point is working on a thesis of a larger research project with some supervision from a professor. There are about 10 regular professors at the school. About half of us are Anglophones and the other half are Francophone and we have support from part-time or adjunct professors as well. Many of them work in the translation industry and then bring their specialized knowledge to share with our students on a part-time basis.
Esther: Can you go through a little bit of your academic background and tell us what drew you to translation studies and technologies in particular?
Lynne: It is probably not an obvious origin story because I grew up in quite a small town in Southwestern Ontario. At my school, we did not have many options for languages, but we are living in Canada so French was offered as a core course for school kids and I enjoyed it. When I started looking at what you could do with studies in French I was influenced a little bit by the older sister of a kid in my class who had decided to go and study translation. At the time I did not even understand that there was a difference between translation and interpretation and I thought, I will be traveling the world and working at the UN and all those kinds of things, so I decided to pursue it. When I got into the program, I realized I am much better suited to translation than interpretation. They have some overlap, but they also have some major differences and I learned early on and quite happily that I am a better translator than interpreter. I also was a student at the institution where I teach now. I did my bachelor’s degree at the University of Ottawa and I had the amazing fortune to meet a professor who was a visionary. Her name was Ingrid Meyer and she has unfortunately passed away but she was the person who brought technology to the School of Translation and Interpretation and at that time technology was word processing. The school was a little bit of a leader. We were doing word processing and terminology management with very simple DOS-based interface tools. When I went to do my work placement, I was out in the field working and the place that I was working at did not even have word processors at that time. They were handwriting their translations and sending them to a typing pool literally. Even though it sounds crazy to think that word processing was cutting-edge technology, it was back in the nineties, so that was my introduction. Even then I could see the potential of technology for making translators’ lives easier.
Professor Ingrid Meyer had a very close collaboration with a professor in the Computer Science Department and she is the one who encouraged me to do graduate studies. I stayed and did my master’s degree at the University of Ottawa as well and I got to collaborate with Professor Meyer in Translation and Professor Skuce in Computer Science and they were working on what they called a knowledge-based approach to terminology management. They had developed a prototype, a knowledge-based term bank and that was my master’s thesis. I was working on part of that project so that was very exciting. I was bitten by the research bug and I wanted to do more. At that time there was no doctoral program in Ottawa in translation and there was nothing in computational linguistics. There was nothing in Canada. You could do linguistics, but not with a strong computational element. I started looking elsewhere and was very lucky to get funding from the Social Sciences and Humanities Research Council in Canada. That allowed me to go to the UK and I did my doctorate in Manchester at what was then the University of Manchester Institute of Science and Technology, which has since merged with the University of Manchester. I continued working for my PhD primarily on terminology and computers but UMIST was the home to the UK part of the Eurotra project, which was an early machine translation project. I got to work there with Harold Somers, who is a big name in machine translation, and some of his other colleagues and so that was my exposure. Manchester had a master’s degree in machine translation specifically. Although I was not on that program, I was still in that crowd. I was mixing with those people and became a big fan of machine translation at that point.
Esther: Tell us about the current Machine Translation Literacy Project that you are working on. In a nutshell, what is the project and why were you inspired to do research in this area?
Lynne: I have spent my whole life in the translation field, but when Google released their first free online machine translation system back in 2006, it was a game changer because it meant that machine translation was all of a sudden in the wild. It was out of the hands of professionals and into the hands of anyone who had an internet connection, so that was huge. What I have started to see is that people with a background in translation bring a very different perspective to using machine translation than people who do not. It is not surprising. Mechanics bring a different perspective to a car than me as a driver of a car. I can use it, but I am not a mechanic. I do not understand all the ins and outs of it and I do not necessarily need to, but there are some things that I do need to know. I need to know how to use the brakes. I need to know how to adjust the mirrors. There are some things that I do need to know to be a good driver and so it occurred to me that these people who have no background in translation, why would they be smart users of it? I wanted this Machine Translation Literacy Project essentially to be an awareness-raising campaign. The technology itself is very easy to use. When we think about technology, we often think that the learning we need to do is a how-to. Which buttons should I push? Where should I click? What order should I do these things in? Machine translation is not like that.
It is a very easy technology to use in terms of copy, paste, click. Sometimes that is a little bit dangerous because we can almost be on autopilot. We do not necessarily think about what we are doing because it does not require a huge cognitive effort to copy, paste, and click. Machine translation is not about how-to, but it is about asking questions like whether or not to. Should I be using machine translation for this particular task? Translators have such a background and it is second nature to them. Evaluating what is the purpose of the text. Who is the intended target audience? All of those things that translators ask themselves naturally, other people do not and so I thought there seemed to be a big scope for including the machine translation component into what we might call just digital literacy. Digital literacy emerged as technology has taken up a larger and larger part of space in our everyday lives and machine translation is one technology that is part of that package now. People have such easy access to it. It is on their phone. They take it with them wherever they go. Sometimes it is even embedded in other tools. They do not even have to consciously use it. It could be invisible almost to them, embedded in their social media. I decided that it is a way for translators to give back and almost a social responsibility. We know how to use these tools and so we should be generous and help other people make good use of them too.
Florian: In one of your YouTube lectures, there is one sequence where you say, usually, people’s perception of MT is on the two extremes. Either it is great and amazing or it is the most ridiculous error ever. Do you have any data? Where does the average consumer lie on that spectrum? Can you give us a sense?
Lynne: I do not have any hard data, but there was an interesting study done by Professor Lucas Nunes Vieira at the University of Bristol and he did an investigation of how machine translation is represented in the popular media. He looked at newspapers over quite a long period and analyzed the presentation of it and he is the one who found that it is quite polarizing. On the one hand, some people are almost thinking we are living in a Star Trek science fiction type of world, and on the other end, there are journalists who portray machine translation as being useless because it cannot do things like translate poetry. We in the translation field know machine translation was never designed to translate poetry. That was not the main driving force behind it, so journalists themselves are not necessarily experts in machine translation, so it is a dominoes effect. The people who are not experts present their view to other people who are not experts and the truth lies somewhere in between. I cannot give you an exact proportion of who believes in which extreme, but we know that those are the principal messages that are out there. There is just a lack of nuance around the discussion of machine translation. Everybody speaks a language and so people feel certain expertise on linguistic subjects that comes just by being a speaker. Perhaps they are speaking a little bit out of turn, not necessarily knowing what they are talking about and then perpetuating some of these skewed perspectives.
Esther: What other members of the team are there on the MT Literacy Project? Have you got other people from the university or department involved and how are you all collaborating if so?
Lynne: We have people who are on the team at the university, primarily students at all levels including the postdoctoral. We have graduate students from both the School of Translation and also from the information school. Then we also work closely with people who are not. This is a project that came out of grant funding so I consider the immediate team the people who are funded with and through the grant. There are colleagues in other universities that are interested in very similar topics and so I have quite a good relationship with colleagues in France, in the Netherlands, and Finland, and so we are working together. With the spirit of wanting to make more resources available, we are all working towards making open resources and collaborating to compile them together. That is the next phase of things. We have all developed our own little things and we are going to be pooling them and making a much bigger collection of resources available. There is a project at the moment in the EU called MultiTraiNMT, which is also about helping multilingual citizens in the European Union become smarter users of MT. There are other projects going on and I do think I have a good relationship with those projects even though they are not technically working on this team. It is definitely a collaborative effort in the wider translation community.
Florian: You recently published on Chinese speakers’ use of MT as an aid for scholarly writing in English and also MT literacy instruction for international business students and business English instructors. Tell us a bit more about those two and some of the findings.
Lynne: My broad goal with the Machine Translation Literacy Project when I started was super ambitious. I was going to help everyone who is not a translator be a better user of machine translation. It did not take me very long before I realized the general public is not one audience. It is made up of many smaller audiences, so I recalibrated a little bit as I cannot help everyone at the same time because different users need different types of guidance. A teenager might need something different than a business student who might need something different than a nurse or somebody else. I decided that I would start with people who were available to me who were on my doorstep and that was international studies. The University of Ottawa has quite a high percentage of international students and so I started working with them. I was offering some small workshops through the library and then I was approached by Teachers of English as a Second Language. They had many international students in their courses and were wondering if we could design something that could help them in particular. That is how I ended up working with those two groups at two different universities. I was working with the business students at Concordia University in Montreal, which is not too far from Ottawa, and the Chinese-speaking students who were at the University of Ottawa.
It was through a second language learning context that I came into contact with them when their teachers reached out. What I found is that the international students who are coming to Canada and need to submit work for their courses to their professors in English, for example, are looking at machine translation as a writing aid. They need to write something for their professor and they are struggling a little bit to write completely independently in English and so they are looking at machine translation as a tool to help them fill the gaps. It is not that they will not even try to write in English, but they do need support that goes beyond what a dictionary might offer for example, so they are looking at it. For these students, an interesting revelation was this idea of garbage in, garbage out. They had no concept of being able to improve the quality of the output text by improving the quality of the input text and this is exactly suited to their need. If they are a native speaker of Chinese, that is their strong language. That is where they have the chance to influence the text by providing a well-written unambiguous source text and it just never occurred to them. They would just write something quickly, feed it through the machine translation system and struggle to patch up the English output in their weaker language. Even this tiny little piece of information which to a translator is second nature, is a revelation to them and can be a game changer for them to make better use of this technology.
Esther: Generally on the use of the free online MT systems, do you see some strong pros and cons in terms of how they are being used?
Lynne: Absolutely. As I said, my first instinct that the general public was one group of users was so wrong. There are so many different use cases out there and some of them are what we might describe as lower risk or lower stakes and then some of them are much higher stakes and that is where there is a lot of work to be done with non-translator users. It is in developing their judgment and risk assessment. They need to stop and ask themselves, what is this text going to be used for? If it is for their own personal consumption, particularly if it is for a hobby or even entertainment? There are lots of kids out there, especially teens who are using machine translation to translate manga comics and anime just for fun, and the stakes there are very low of getting a poor translation. The worst thing that is going to happen is that you are going to be disappointed, you did not understand the manga comic. It is not life or death.
Then we also hear some stories COVID-19 has brought to the forefront about health agencies using machine translation to communicate with the public and some mistranslations. There was one example about a health agency in the United States who had used machine translation on their website. They were in fact trying to encourage people to get vaccinated, but the text was into Spanish and it was mistranslated. Instead of saying, the vaccine is not mandatory or not required, what it ended up seeing was that the vaccine is not necessary and that is a completely different message and there are some real consequences to that. The places where it is not a good idea to have raw machine translation, it does not mean there is no role for machine translation, but we do not want to have a raw machine or unedited machine translation when there are risks to people’s health. We have also heard some very difficult stories about machine translation being used in immigration situations and somebody’s life is hanging in the balance. You are making a huge decision about someone’s future. You do not necessarily want to trust that to raw machine translation so there is education to be done about higher stakes and lower stakes tasks and where we can use it without being too worried and where we need to give it a second thought. As I said, it does not mean there is no role for machine translation but maybe we are not going to go with just the raw unedited machine translation in those higher-risk situations.
Florian: What do you think about confidentiality and privacy? For example, we recently covered this story about Swiss Post blocking and then unblocking DeepL because of privacy concerns. They kept it blocked for the banking arm, Post Finance, because of confidentiality and regulatory concerns. What are your thoughts on using these free online tools?
Lynne: It is something that people are surprised to learn. Most people and myself included do not read the terms and conditions of every single app and program and site that I go to. A lot of us click through, accept, and do not read them. It is one of these things that is a revelation to people outside the translation industry that free online tool providers often have in their terms and conditions something that says they are able to keep your data. They are able to reuse it or repurpose it for continued training of the machine translation system or maybe for other things. A lot of the things that everyday users use machine translation for are not highly confidential, but it only takes one or two slips. Maybe it is health information that you put in there. Maybe it is banking information. When I have shared this with students, their initial reaction is shock horror and then they shake their head and say, now that I think about it I am not surprised, I probably should have realized that, but they do not so it is worth that nudge. Use your judgment. Think twice. You can enter a lot of text and still leave out the critical bits so that you can still get a lot of the text translated but you are not giving away the essential, private information. In the translation industry, it is different but that is not the community that I have been targeting with this project. There is some awareness-raising to happen within the translation community as well about client confidentiality because in those cases it is not necessarily personal information. With users outside the translation community, they are more likely to share personal, sensitive information. When it is within the translation community, the client has different reasons for keeping what they consider confidential information that translators need to be aware of. That is an example of where machine translation literacy is important for different people, but different people need different types of information as part of that literacy training, so it is not like MT literacy can just be one thing. There are going to be some core elements that are relevant to everyone, but there is also some customization that needs to happen for different user groups.
Florian: There is this balance now that you need to teach the core skill of language competency and domain expertise in certain areas that they want to maybe pursue a career, but then there are also the basics of Python or how neural networks work. You do not want to overload it by having half of the curriculum being a Python program. How do you balance that?
Lynne: It is getting harder because everything that you add to the curriculum usually means taking something else out because we still have a fixed period of time. We are not all of a sudden offering a ten-year BA because nobody would sign up for that. It is getting harder. We have seen that some of the things that we used to have to literally teach like word process we do not have to do anymore. Some things have become more part of our general knowledge, general culture but other things are being added all the time and Python is a great example. How we are handling it for the moment is that we are teaching what we see as core technology tools, so we are looking at terminology management tools, translation memory tools, machine translation, concordances, and especially how all of those can work together in a suite of tools. Python programming at the moment is offered, but it is not compulsory, so for students who are tech-oriented and want it, they can get it, but it is not something that we are building into our core program. We have collaborations at our institute. I do not think it is unique to us, but at the University of Ottawa, there is a program in Digital Humanities and they are ones who offer the Python programming courses and we would recognize that as an elective that could count towards our program for a student who wanted to do that.
Esther: How was all of the teaching happening during lockdowns? Did that change what you were able or what you wanted to do in Translation Studies? It might have had more of an impact on the conference interpreting, but what adjustments did you see?
Lynne: We all had to make a pretty quick transition everywhere around the world from in-person to online. We allow people who do not want to do the full BA to still get a certificate, for example, and we were already offering that online so that was an advantage. We already had some courses that had specifically been designed for online so that was good and some professors knew how to do that. Other courses had been hybrid which meant some weeks were in person and other weeks online so they were partly online already. That was a real advantage because we were not starting from zero but there was still a lot of work to do. We are very lucky now that we do not have to install many of the tools locally in our computer lab. We can access them through the cloud, so that is a huge advantage. If this pandemic had happened even five years ago, we would have struggled so we are lucky to have these cloud options now which allowed a seamless transition at least from that respect of accessing tools and being able to use them.
Florian: To end on a goal and next steps note for the MT Literacy Project. What are the next steps or project goals?
Lynne: I have been convinced by working with first-year university students that this information could be valuable to them even in secondary school. That is one of the things that I want to do next and I am trying to think about a smart way of doing it. Me as an individual or even my team as half a dozen individuals can only reach so many people directly and what I am thinking now is that this MT literacy does not need to be a standalone package. It fits in well with digital literacy and information literacy and these are subjects that are taught in high school, so kids in high school or secondary studies are learning digital literacy skills and learning information literacy skills. What I am hoping to do next is to connect with teacher training here at the University of Ottawa so we have a faculty of education and I am hoping to be able to train the trainers. If I could get in front of the people who are training to become teachers and get machine translation literacy incorporated in that digital literacy curriculum, then I could teach the teachers and the teachers would go on to reach thousands of students at the secondary level. That is a much smarter way.
I would also encourage other people in the translation community to be doing similar things. Even sharing our infographics. It is not difficult knowledge to share. It is not revolutionary certainly to people in the translation industry and it often does not take much for people outside the translation community to see the light bulb go off. Just this simple thing like telling them not to put your personal information in there. Do you not know that that information can be kept and shared? Or garbage in, garbage out. They are fairly simple concepts because we are not trying to train people to become translators. We are not trying to give them translator knowledge. What we want is for them to just realize you can use this technology in a slightly smarter way and see benefits, so it is not like competing with translators. These people are not interested in becoming professional translators. That is not their goal either. They just want to make better use of a tool that is available to them. Training the trainers and working with teachers is one big thing and I do think there are other ways of connecting with teens as well. I am a big fan of libraries. I have a cross-appointment to the library school, so I have been working mostly with academic libraries right now to reach students at universities but the public library is another venue that has enormous potential for reaching families, kids, teens. I would also like to do more work moving forward with the public library as another group that is open to sharing information and supporting people in the community.