After a rollicking start to our Beat the Machine online translation slam, we’re back with a new sentence! If you need a quick refresher, you can read about the premise in our inaugural post here. Very simply, we’re out to prove how much better human translators are than machines and maybe learn something from one another in the process.
Here is the sentence for this month:
Mais si les relations sont aujourd’hui plus conflictuelles que jamais, il était pour ainsi dire écrit dans le ciel que la formidable croissance économique chinoise des quarante dernières années, orchestrée qui plus est, depuis huit ans, par un régime Xi particulièrement autoritaire et expansionniste, finirait par déboucher sur une lutte de pouvoir de grande envergure entre la Chine et un empire américain qui n’est forcément plus ce qu’il était.
What was that I said last month about French being fond of long sentences? This one will give you ample opportunity to wade through and potentially break into as many shorter sentences as you see fit. The sky is the limit!
For context, Xi refers to Xi Jinping, the President of the People’s Republic of China, who has been in power since 2012. His name doesn’t require any particular treatment, and “Xi regime” would be a fine translation in this context (but feel free as always to get creative!).
Here is the full paragraph the sentence came from:
Que Pékin ne joue pas franc jeu en matière commerciale est notoire et que la dictature chinoise ait depuis longtemps à l’ordre international un rapport « inadéquat » est incontestable. Que M. Trump joue la corde antichinoise à l’approche de la présidentielle, il fallait s’y attendre. Mais si les relations sont aujourd’hui plus conflictuelles que jamais, il était pour ainsi dire écrit dans le ciel que la formidable croissance économique chinoise des quarante dernières années, orchestrée qui plus est, depuis huit ans, par un régime Xi particulièrement autoritaire et expansionniste, finirait par déboucher sur une lutte de pouvoir de grande envergure entre la Chine et un empire américain qui n’est forcément plus ce qu’il était. Nous y voilà. Pour l’heure, l’ordre du monde est façonné par les faucons des deux côtés.
If you’d like to read the full article from Le Devoir, you may find it here.
Here is Google’s feeble attempt:
But if relations are today more conflictual than ever, it was almost written in the sky that the tremendous Chinese economic growth of the last forty years, orchestrated moreover, for eight years, by a particularly authoritarian and expansionist Xi regime , would eventually lead to a large-scale power struggle between China and an American empire that is not necessarily what it used to be.
Submit your much better translation hereby September 30, 2020, and the blog post discussing it will go live in October!
Please note the following:
Only FLD members will have their translations posted on this blog. Membership is free for current ATA members, so if you aren’t a member yet, make sure to join before you submit your translation!
You are free to submit your sentence anonymously, but half the fun will be crediting the creative submissions we receive by name and recognizing their authors.
You may submit as many times as you like in case you have a stroke of genius after your initial submission. I will only discuss one submission per person in the review post.
Have you translated or read a particularly pesky sentence this year that you can share for this project? Please send it along! Are you interested in helping us do the same virtual translation slam, but from English to French? We’d love to have one or more volunteers to do this series, but in reverse! If you’re interested, please contact Ben Karl, the À Propos editor, at ben [at] bktranslation.com or myself, Sam Mowry, at sam [at] translation.expert to let us know!
Welcome to part two of our inaugural Beat the Machine mini translation slam! In our July post, I asked FLD members to re-translate a complicated sentence to improve upon the machine translation output provided in the post. Now it’s time to review some of the submissions!
Before diving in, I would like to acknowledge that I may have had a little bit of hubris when selecting this first sentence. One of my favorite FLD members suggested it and I thought, “perfect, this thing is a mess, let’s do it!” Then I sat down to tackle it myself and… eesh, this sentence really was a challenge! The good news is that this has been a fun learning experience for us all, and I now know what I’m looking for in future sentences. The other good news is that many of our colleagues were clearly less daunted than I and submitted some really creative solutions. So, let’s dig in!
To refresh our memory, this was the sentence we were working with:
L’excellentissime pianiste classique autrichien Friedrich Gulda n’eût peut-être pas été d’accord, lui qui ne cessa de transgresser les deux grands ordres (jazz et classique) en les reprisant et déprisant dans des concerts qui filaient standards de jazz, classiques des classiques.
And here’s what DeepL gave us:
Perhaps the excellent Austrian classical pianist Friedrich Gulda would not have agreed, as he never stopped transgressing the two great orders (jazz and classical) by reproducing them in concerts that spun jazz standards, classics from classics.
As foreshadowed, there are a lot of things we’re working with here. There are four different parts to this monster, so let’s take it part by part before addressing some really neat things some people did with the structure of the sentence as a whole.
L’excellentissime pianiste classique autrichien Friedrich Gulda…
In the first seven words, everyone agrees about three of them. “Friedrich Gulda” and “Austrian” are pretty concrete! Things immediately diverge after that. Some of the options for excellentissime were excellent, brilliant, and outstanding. However, two people did something pretty clever here, opting for “virtuoso” in English. This is particularly delightful because it folds the level of skill into the noun: Friedrich Gulda, classical piano virtuoso. The alternative, [adjective] + [classical pianist], is perfectly accurate, but virtuoso conveys a level of talent beyond “excellent” that better matches excellentissime (the –issime meaning very excellent) and changes up the sentence structure ever so slightly.
…n’eût peut-être pas été d’accord…
Everyone went with either “would not have agreed” or “would have disagreed.” This a fun reminder that you can structure even seemingly straightforward text more than one way. The difference is slight, but real, and which option is “better” depends on the rest of the sentence: is the goal fewer total words? Shorter words? Depending on the context, choosing something like “may have begged to differ,” could potentially be great.
…lui qui ne cessa de transgresser les deux grands ordres (jazz et classique)…
Here’s where things start to get messy. Ne cessa de became: ceaselessly, always, continued to, constantly, and continually. So many options to convey “something that never stops!” Transgresser is clearly a problem in English, as “to transgress” is much weightier than just mixing musical genres, not to mention the moral or religious overtones. DeepL fell right into this trap. Our human options here included “intermixed,” and “transcended.” My favorite solutions were “went outside the box,” which, while it could use a stronger verb than “went,” encompasses the notion of transgressing in a more palatable way, and my very favorite, “pushed the limits.” He didn’t necessarily break the boundaries, as a transgression might suggest, but he’s right up against it.
…en les reprisant et déprisant dans des concerts…
Oh no, wordplay in the source! The holy grail here would be to come up with something that has the same kind of parallelism or at least some kind of interplay as reprisant et déprisant. Options included: “combining and undoing them,” (accurate, if not a little clunky), “reappraising and transforming,” (yet clunkier, in my opinion). “Taking them apart and putting them back together”: we’re getting there, it’s literal but closer to the mark. “Deconstructing and reconstructing,” is almost there and is the best non-metaphorical option that was submitted; it checks both boxes, opposite words with a similar structure to match the source.
However, there was one superlative submission here that does all of the above but also leads into the next part of the sentence beautifully: “unraveling and reweaving them in concerts.” I swooned.
…qui filaient standards de jazz, classiques des classiques.
The swoon-worthy submission continues: “…that spun together jazz standards and classical classics.” This is why I love this option so much for the previous part. The translator saw filer in this section and put it to excellent use in the previous one, using a thread metaphor to describe how Gulda took apart and reassembled the musical components. The use of “spun” continues the metaphor perfectly.
Lastly, “classical classics,” submitted by two translators, is snappy and alliterative, and I don’t know what more you could ask.
Sentence breaks? What are those?
This mouthful of a French sentence reminds us that what is valued for style in French doesn’t always correspond to what we look for in English: French sentences can run on and on and on. And I actually chopped this sucker in half before issuing this challenge! Two of the translators used a period and broke it into two fully separate sentences; one person used a semicolon for the same purpose. One option that surprised me was to pull the initial verb (“might have disagreed”) all the way to the end, so it read something like “Friedrich Gulda, [description], who [did the things], may well have disagreed.” This, again, is a decision where context matters, and this option may or may not flow into what comes next in the text. But it certainly has the option to, and that’s awesome. Next time, I’ll provide more context so that we can better evaluate options like this one.
Putting it all together, this is a string combining my personal favorite individual translation solutions for this sentence:
Friedrich Gulda, the classical piano virtuoso who continually pushed the limits of two great genres (jazz and classical), unraveling and reweaving them in concerts that spun together jazz standards and classical classics, may well have disagreed.
You know what? I think that’s pretty good! That’s a well-crafted sentence. And, given the notable lack of preceding or following sentences, I can claim it is ideal for the context. So… whew! We made it! I hope that it was useful and informative to see how many options there are for even simple phrases, and what neat things you can achieve with even mundane words. Stay tuned for next month, where we’ll do it all again (with a more approachable sentence this time!).
Did you forget to submit a translation in time? Not to worry! Share your version on Twitter and tag the French Language Division (@ATA_FLD) and me, @SamTranslates.
If you would like to submit a sentence for a future slam, I would like that very much! You can contact me, Sam Mowry, directly at sam [at] translation.expert or on Twitter at the handle listed above. You can also contact the À Propos Editor Ben Karl at ben [at] bktranslation.com.
If you’d like to help launch a similar slam but into French, please also reach out!
Beat the Machine: A Mini Virtual Translation Slam by the ATA FLD
Greetings, fellow FLD members (and interested onlookers)!
My name is Sam Mowry and I’m here to help us all translate better. Strong words, I know, but one thing I’ve found to be true in my translation career is that the more exposure you have to translations, particularly good translations, the better translator you become. To that end, this is the first in what I hope will become an ongoing series of posts here in À Propos.
The premise is simple: I’ve never met a translator who, when confronted with someone else’s translation, doesn’t secretly or not-so-secretly think to themselves, “I could have done it better.” Moreover, as human translators, we know we’re vastly superior to every machine translation option on the market. We’re going to combine those concepts into a monthly “beat the machine” virtual translation slam (and by that I mean slamming those machine translations into the ground!).
Every two months, I will post a French sentence with an English translation produced by a widely available machine translation engine. This will incite the faithful readers of this blog to rise to the challenge and show how much better it could be by submitting their own versions of the translated sentence. The following month, I will publish a blog post where I share some of the best submissions and discuss what makes them so good. This is a chance to show what a difference the human touch makes and improve our own translation practices in the process by seeing how other translators approach the same problem.
Sound good?
The first sentence is:
L’excellentissime pianiste classique autrichien Friedrich Gulda n’eût peut-être pas été d’accord, lui qui ne cessa de transgresser les deux grands ordres (jazz et classique) en les reprisant et déprisant dans des concerts qui filaient standards de jazz, classiques des classiques.
Fun, right? Hat tip to FLD member Beth Smith, who provided this sentence. Here is what DeepL spat out:
Perhaps the excellent Austrian classical pianist Friedrich Gulda would not have agreed, as he never stopped transgressing the two great orders (jazz and classical) by reproducing them in concerts that spun jazz standards, classics from classics.
You are no doubt chomping at the bit already to submit your much better translation of this sentence. You can do that HERE.
Submissions must be received by July 22, 2020. The follow-up blog post discussing the best solutions will be posted on or around August 1, 2020.
Please note the following:
Only FLD members will have their translations posted on this blog. Membership is free for current ATA members, so if you aren’t a member yet, make sure to join before you submit your translation. When you log in to your account on the ATA website, the number of divisions you belong to is listed at the top of the page. Click “Modify” to change which divisions you belong to (and add the FLD!).
You are free to submit your sentence anonymously, but half the fun will be crediting the creative submissions we receive by name and recognizing their authors.
You may submit as many times as you like, in case you have a stroke of genius after your initial submission. I will only discuss one submission per person in the review post.
Finally, we’re hoping to continue this series with all of your help! Have you come across a particularly pesky sentence you can share for this project? Please send it along! Are you interested in helping us do the same virtual translation slam, but from English to French? We’d love to have one or more volunteers to do this series, but in reverse! If you’re interested, please contact Ben Karl, the À Propos editor, or myself, Sam Mowry, to let us know!
Happy translating!
Sam Mowry is an ATA-certified French into English translator specializing in international development, medicine, official documents, and being mouthy on the internet. She can be reached by email at sam@frenchtranslation.expert or directly on Twitter at @SamTranslates.
Welcome to the 16th episode of the French Language Division’s Continuing Education Series. In today’s episode, Jenn Mercer (FLD Administrator) and Andie Ho (FLD Assistant Administrator) welcome our new Podcast Host, Cathy-E Nzume!
Cathy-Eitel Nzume is a certified French to English and English to French Court Interpreter, translator, Department of State Certified Linguist and Legal professional. She holds a Master’s degree in International and American law from Howard University, as well as a Master’s degree in European and International law from the University of Amiens, France. She grew up in her beloved country of origin, Cameroon that she left to study abroad. She specializes in legal and conference Interpreting as well as legal and financial translation. You can find her on LinkedIn at https://www.linkedin.com/in/cathyeitelnzume/ or on Twitter at https://twitter.com/cathyenzume
HOW TO LISTEN
SOUNDCLOUD: You may listen to or download Episode 16, and our previous episodes, on Soundcloud here.
ITUNES: This episode, and our previous episodes are available on iTunes here. You can subscribe or listen online.
For SoundCloud and iTunes:
Welcome to the 16th episode of the French Language Division’s Continuing Education Series. In today’s episode, Cathy-Eitel Nzume welcomes Jenn Mercer and Andie Ho to discuss the State of the FLD and the upcoming ATA Conference that may or may not be in Boston. For more information about our podcast, please visit the French Language Division’s website here: www.ata-divisions.org/FLD/category/podcast.
A veteran SAMiste shares her experiences at the 2018 Medical English Seminar in Lyon—will you join her in 2020?
By Anne-Charlotte Giovangrandi
Photo Credit: Indelebile photographe
This article was originally published in Translorial, the journal of the Northern California Translators Association (vol. 41, no. 2, fall 2019) and is republished here with permission.
Short for Séminaire d’anglais médical, SAM is organized every other year in Lyon, France, by the Société française des traducteurs, the French sister association of the ATA. Presented as a medical English writing and terminology training course, SAM is geared toward translators working in French and English who specialize—or wish to specialize—in medicine. It attracts linguists from all over the world, most coming from France and the UK. This article reviews the 2018 conference, which was held over five days at the medical school of Université Claude Bernard Lyon 1. As registration for the 2020 conference will open early next year, now seems to be a good time to spread the word!
In March 2018, participants (nicknamed “SAMistes”) who were lucky enough to arrive in town a bit early attended a welcome cocktail party on Sunday night. Newcomers were given the chance to start getting to know other participants over a glass of wine and plates of local charcuterie and cheeses, while old friends from previous conferences greeted each other joyfully, happy to reconnect.
Lectures and terminology sessions
The conference format alternates medical lectures (presented in French or English by expert guest speakers) with related terminology sessions led by Nathalie Renevier, a renowned translator, translation instructor, and terminologist who has translated numerous health and medical publications. The 2018 conference presentations covered a wide range of subjects: acute medical care, developmental coordination disorder, type 2 diabetes, the MeSH thesaurus, schizophrenia, dermatology, PTSD, and European Medicines Agency templates. In 2016, the presentations were just as diverse, including sessions on Alzheimer’s disease, influenza viruses, statistical analysis for clinical trials, and three cancer-related presentations.
All attendees can benefit from the lectures, which are presented simply enough to appeal to newbies but nevertheless provide enough detailed information to satisfy the intellectual curiosity of seasoned members of the field. The terminology sessions, in which audience participation is encouraged, are invaluable for our community of linguists. The organizers facilitate the learning process by providing all presentations, including all the terms discussed during terminology sessions and many valuable references, shortly after the event. Personal medical glossaries no doubt expand tremendously by the end of the workshop!
Hands-on learning
The first afternoon of the week is always spent working hard: the SAMistes are split into small groups to translate a 350–500-word excerpt from a selection of medical texts. In 2018, possible passages featured subjects as varied as genetic mutations, lip physiopathology, bulimia, and schizophrenia. The texts are sent before the conference, and most of the participants read them in advance; some even manage to prepare a draft translation to use as a starting point. Working in groups is not common in our profession, so this exercise offers participants a rare opportunity to glimpse colleagues’ methods for tackling a translation task, share tips and favorite resources (both paper and online), and brainstorm the perfect term or idiomatic expression to produce the best collaborative target text.
Over the course of the week, each group presents their translation to all the participants, with one of the guest speakers or organizers answering questions and providing feedback.
The extras
SAM offers multiple opportunities to network, chat with colleagues, and have fun, whether during the coffee breaks, over lunch at one of the nearby restaurants, or while attending the specially organized gourmet dinner. In 2018, attendees also enjoyed a guided tour of the fascinating exhibit Venenum: A Poisonous World at the architecturally stunning Musée des Confluences (https://xl8.link/204).
Some fun to conclude the hard work
In what has now become a tradition, SAM concluded on Friday afternoon with two translation slams (one for each translation direction) between two translators brave enough to each present their own translation of the same satirical scientific article. If you want an idea of the type of humor, check out the Onion article chosen for the 2018 English-to-French slam (https://xl8.link/203).
Attendance has grown steadily over the years; 2018 saw 60 participants, including the organizers and the terminologist. Almost three fourths of them were returning SAMistes. No wonder, as the format is quite addictive, as are the culinary delicacies of the host city! As a mise en bouche, the full program for 2018 is available at https://xl8.link/206.
An English into French translator since 2010, Anne-Charlotte Giovangrandi moved to the San Francisco Bay Area from her native Switzerland almost 20 years ago. Holding a translation certificate from New York University and a master’s degree in psychology from the University of Lausanne, she has developed a double specialization, translating patient-facing medical documents and transcreating marketing/advertising content (high jewelry in particular), an area in which she can let her creativity run free.
While most of America was grilling out, drinking beer and enjoying the fireworks on the 4th of July, I was sitting in a conference room learning about hedging foreign exchange (FX) risk, the transition from LIBOR and EURIBOR to alternative interest rates, BASEL 3 capital requirements and interest rate benchmarks, and participating in workshops on how to edit for the European Central Bank and write clear corporate communications, among other things. I plan on spending my 4th of July holiday the same way every year.
Spiez, Switzerland. Photo credit: Amanda N. Williams
Why on earth would I do this to myself?
I’m sure we’ve all heard one version of this or another, but over the past ten years, our market landscape has started to change. New technologies are being introduced every day, translation companies are consolidating into mega-companies, some translators are experiencing pricing pressures—the market is fracturing, and we’re all trying to find our rightful place and still make a good living in this new world. For me, finding my place involves constant work to improve my subject matter expertise, target language writing skills and source language skills. I want my work to stand out. That’s why I make it a priority to attend conferences overseas every year.
Portfolio table, ASTTI conference 2019. Photo credit: Martin Hemmings
But can’t you do all those things in the US? Not nearly to the same extent. I’ve found that conferences in Europe tend to be much more specialized and in-depth. They’re often in your source language country (although that wasn’t the case for this Swiss conference I just attended), and I’ve noticed that at European conferences, translators are more open to showing their work (in portfolios and in workshops), so you get to see some really outstanding translations. It really opens your eyes to a whole new world of possibilities and pushes you to take your own work to the next level.
ASTTI Financial Translation Summer School
Nestled in the Bernese Highlands of Switzerland on the shores of Lake Thun, Spiez is a picture-perfect Swiss city that looks like it came straight out of The Sound of Music. You couldn’t pick a better location for a highly specialized translation conference. On July 3, 2019, over 80 financial translators from 10 countries as far away as the United States and Australia came together to attend this three-day deep dive into finance and financial translation, organized by the Association suisse des traducteurs, terminologues et interprètes (ASTTI).
ASTTI session in progress. Photo credit: Rosie Wells.
We spent our days attending incredibly technical sessions and rigorous three-hour hands-on workshops and our evenings cruising the pale blue waters of Lake Thun or relaxing with a pint at the quirky and quaint Piratenbar, hidden on a back street near the lake. Downtime is practically required after concentrating so hard all day, and the conference organizers managed to strike a perfect balance for attendees. Here are some of my favorite highlights.
Translations are like fingerprints – no two are alike
This year, ASTTI hosted its very first translation slams—French to English, English to German, English to French and English to Italian. The French to English slam pitted Martin Hemmings, MITI (our distinguished speaker at ATA60 in Palm Springs, by the way) against Sylvia Smith. It was one of the most fascinating slams I’ve ever seen.
French to English translation slam: Martin Hemmings vs. Sylvia Smith. ASTTI conference 2019. Photo credit: Michael Dever
The ASTTI organizing committee did a fantastic job selecting these slam participants. They are both talented translators with impressive portfolios in their own right, and their respective styles couldn’t be more different. Sylvia works for a Swiss bank and has a very technical background, while Martin mainly works for big NGOs, government agencies, and corporations looking for target audience-focused translations. Both translations were well-written, but completely different. Sylvia took a careful banker’s approach, adding in explanations where she thought necessary, while Martin assumed the target audience had the appropriate background knowledge to understand the text and focused on style and concision. Sylvia’s translation ended up being longer than the French while Martin’s was 20% shorter.
That slam was the clearest example I’ve ever seen of how two exemplary translations of the same text can be so utterly different.
And for those who have not yet met Martin or seen his work, you don’t want to miss his two workshops at ATA60 in Palm Springs. Trust me on this. I squealed with joy when I learned he was coming to present at the ATA conference.
Getting into the nitty gritty
As I mentioned above, this conference was highly technical. Full hours were devoted to specific topics in finance. I love this because the more we know about a subject, the better we’re equipped to translate about it. I dealt with accounting in my former career, but the world of accounting and finance is so vast there’s always some aspect of the field you can learn more about. And things change all the time, so it’s important to get refreshers and updates on a regular basis.
Prof. René Sieber discusses FX risk management at ASTTI 2019. Photo credit: Rosie Wells.
View of Lake Thun from Hotel Belvedere. Photo credit: Amanda Williams.
Lake Thun. Photo credit: Rosie Wells.
ASTTI 2019 dinner cruise on Lake Thun. Photo credit: Rosie Wells.
ASTTI 2019 workshop. Photo credit: Rosie Wells.
Lake Thun. Photo credit: Rosie Wells.
Marcus Littler talks about money market funds at ASTTI 2019. Photo credit: Rosie Wells.
Paul Bakker talks about financial crisis reforms at ASTTI 2019. Photo credit: Rosie Wells.
Lake Thun. Photo credit: Amanda Williams.
Sandro Graf discusses behaviorial lessons for communication design at ASTTI 2019. Photo credit: Rosie Wells.
For the most part, the presenters of these sessions were not translators—they were bankers, university professors, a former CEO of the Belgian Stock Exchange and current macro-economist, a member of the European Central Bank, etc. Who better to teach us about current trends in accounting and finance than the people living them?
“The most important investment you can make is in yourself” — Warren Buffet
Sunset dinner cruise on Lake Thun. Photo credit: Amanda N. Williams
Is it expensive to attend conferences in Europe every year? Yes. Is July 4th weekend the most optimal time to go? Not exactly. Will these things deter me from going? Never, because what I gain from these conferences is invaluable—new leads, new colleagues and friends, a whole lot of knowledge and a thirst to constantly improve. Every year I attend, I’m a little less afraid of a machine taking my job as AI and NML progress, and that peace of mind is worth its weight in gold.
Photo credit: Amanda N. Williams
Amanda N. Williams is an ATA-certified French to English translator specialized in business, international trade and financial translation. Prior to becoming a translator, she had a career in international trade where she held roles in sales, operations and trade compliance management.
Amanda currently serves as assistant administrator for the ATA’s Literary Division. She also served six years on the Board of Directors of the Atlanta Association of Interpreters and Translators (the Georgia Chapter of the ATA). You can find her on Twitter as the Adorkable Translator (@Adorkable_Trans), on her website at www.mirrorimagetranslations.com or you can reach her via email at amanda@mirrorimagetranslations.com.
FRENCH LANGUAGE DIVISION DINNER AT ATA’S 60TH ANNUAL CONFERENCE
The French Language Division’s dinner at the 2019 Palm Springs conference will be held at Eight4Nine. We hope to see you there! WHEN
Friday, October 25, 2019 at 7:00 p.m. WHERE
Eight4Nine
849 N. Palm Canyon Dr.
Palm Springs, CA 92262
(760) 325-8490 https://eight4nine.com PAYMENT AND RESERVATIONS
Price: $63.00 per person. This includes three-course dinner (see menu below), tax, and gratuity. NOTE: All non-alcoholic and alcoholic beverages will be the diner’s responsibility and are *not* included.
Please select the “send money to friends and family” option so that the FLD is not charged additional PayPal fees.
IMPORTANT NOTE: We do not provide refunds. You may sell or give your ticket to another conference attendee to attend in your place. If you do so, please notify us of the update, but the FLD does not provide refunds once a spot for the dinner has been purchased.
TRANSPORTATION
Eight4Nine where the FLD dinner is being held is located 1.2 miles (approximately 25 min walk) from the conference convention center. As this is farther than previous dinners, you may prefer to use Lyft or Uber.
MENU (you will choose your dishes the night of the event) APPETIZER
Simple Green Salad Hand-selected baby lettuces, fresh herbs, lemon vinaigrette, and rose petals. Gluten-free.
Avocado and Roasted Tomato Tart (vegetarian) Vine-ripened cherry tomatoes, house-made ricotta, black garlic butter, fresh herbs, cracked pepper, toasted brioche ENTRÉE
Cauliflower Steak and Purée Walnut-caper salsa with Fresno chili, dried currants, served with lemon wedges. Gluten-free, vegan.
Grilled Scottish Steelhead Fennel, orange, basil, crispy rice cake and arbequina olive oil. Gluten-free.
Beef Tenderloin All-natural filet of beef, St. Agur blue cheese and red wine reduction, fingerling potatoes, roasted carrots. Gluten-free.
DESSERT
Peach Cardamom Crème Brûlée
Gelato or Sorbet
Note: Drinks are not included.
===================================================================
Welcome to the 15th episode of the French Language Division’s Continuing Education Series. In today’s episode, André Racicot joins Angela Benoit to discuss the translation of toponyms.
André Racicot is a retired English to French translator, editor, terminologist, and trainer from the Translation Bureau of the Government of Canada. He holds a master’s degree in political science and a certificate in German studies. He focused on translation of foreign geographical names. He published a List of Names for Countries, Capitals and Inhabitants in 2000. This list was integrated into the style guide of the Canadian Department of Foreign Affairs. You can find him on Twitter at @AndrRacicot or contact him through his website.
HOW TO LISTEN
SOUNDCLOUD: You may listen to or download Episode 15, and our previous episodes, on Soundcloud here.
ITUNES: This episode, and our previous episodes are available on iTunes here. You can subscribe or listen online.
For SoundCloud and iTunes:
Welcome to the 15th episode of the French Language Division’s Continuing Education Series. In today’s episode, André Racicot joins Angela Benoit to discuss the translation of toponyms. For more information about our podcast, please visit the French Language Division’s website here: www.ata-divisions.org/FLD/category/podcast.
Le présent article est directement inspiré d’une présentation orale donnée (en anglais) à l’occasion du Congrès annuel de l’American Translators Association, le 25 octobre 2018, à La Nouvelle-Orléans, Louisiane, USA. Le titre original de cette intervention était : « Don’t Trust the Machine : How Neural Persuasive Power Can Mislead Post-editors ».
L’arrivée de la traduction automatique neuronale (NMT) a provoqué en 2017 une mini-onde de choc dans le monde de la traduction professionnelle. Pratiquement du jour au lendemain, les traducteurs ont pris conscience que la machine pouvait, à tout le moins dans certaines combinaisons linguistiques et certains domaines, produire du contenu « utilisable ». Cette évolution a parallèlement alimenté bon nombre de fantasmes et de craintes dans le chef des « consommateurs » et des prestataires de services de traduction.
Dans le giron des associations professionnelles ou sur les réseaux sociaux, les réactions ont été – et sont encore – aussi nombreuses que passionnées. Entre commentaires alarmistes, jugements infondés et exemples confinant à l’absurde tirés de textes de Shakespeare, on constate que les professionnels de la traduction ne sont pas toujours bien informés des tenants et aboutissants de la traduction automatique et de la post-édition et que les avis émis, guidés par l’émotion, pèchent par manque d’information et de données objectives.
L’objectif de la présentation donnée lors du Congrès annuel de l’ATA en novembre 2018 à La Nouvelle-Orléans était de dépasser les attaques ad machinam que lancent les militants anti-MT sans véritable fondement et de mettre le doigt, à travers des données très concrètes, sur les avantages et les dangers du recours à la (post-édition de) traduction automatique dans un contexte professionnel. Ma présentation s’articulait en trois parties, avec une introduction sur la technologie (histoire, évolution, perception des différents acteurs du marché), la présentation des résultats de deux enquêtes menées à l’échelle européenne sur l’utilisation de la technologie, et l’exposé d’une étude de cas réalisée auprès de professionnels et d’étudiants.
Introduction
L’introduction retraçait brièvement l’histoire de la traduction automatique et présentait mon opinion personnelle sur la perception de la technologie par les différents acteurs du marché. J’ai ainsi expliqué, non sans une petite dose d’ironie et d’autodérision, les quatre « profils » que l’on retrouve sur le marché des traducteurs indépendants : 1. les partisans de la politique de l’autruche, adeptes des exemples shakespeariens évoqués plus haut, 2. les fatalistes, persuadés de la mort imminente de leur profession, 3. les fourbes, qui critiquent ouvertement la traduction automatique en public, mais qui l’utilisent en secret dans la quiétude de leur bureau (soit par facilité, soit par obligation), 4. l’utilisateur réfléchi, qui se pose systématiquement la question de la pertinence, en fonction du contexte d’utilisation (profil recommandé).
Il convient aussi de noter que les opinions sont très divergentes en fonction de la position des acteurs sur le marché. Un grand nombre de pratiques existent, allant des arnaques pures et simples des agences low-cost, qui tentent régulièrement de faire passer des textes traduits automatiquement pour des productions humaines, aux agences plus consciencieuses qui essaient de prendre le train en marche sans effrayer leurs précieuses ressources linguistiques, en passant par les étudiants et les formateurs en traduction, dont les pratiques d’apprentissage et d’enseignement sont bouleversées, doucement, mais sûrement, par l’arrivée de la technologie.
Ces constats indiquent qu’il règne une extrême confusion sur le marché, que ce soit parmi ses acteurs (traducteurs, clients, tiers) qu’au sujet du produit lui-même (confusion entre qualité du contenu brut et qualité du contenu post-édité, etc.). De cette confusion découle un grand besoin d’objectiver le débat et de pédagogie.
Quelques statistiques
La deuxième partie consistait à dispenser quelques données objectives sur l’utilisation réelle de la traduction automatique au sein du secteur. Pour ce faire, j’ai présenté les résultats de deux enquêtes menées en 2017 et 2018.
La première, « European Language Industry Survey », menée en 2017 par un ensemble d’organisations (Commission européenne (via le réseau EMT), Elia, GALA, EUATC et FIT Europe), indiquait entre autres que l’année 2017 avait été la toute première année dans l’histoire de la traduction où la proportion de professionnels (agences et traducteurs individuels) affirmant avoir déjà eu recours à la post-édition de traduction automatique avait dépassé la barre symbolique des 50%.
L’autre enquête présentée, menée par la Chambre Belge des Traducteurs et Interprètes en 2018, indiquait une réticence plus marquée des traducteurs individuels face à la technologie, mais avec un « potentiel d’évolution » non négligeable lié à la réponse « I might use it in the future ». Ces deux enquêtes montrent un recours exponentiel à la post-édition de traduction automatique par tous les acteurs du marché. Une tendance qui devrait se poursuivre à l’avenir.
Étude de cas
Après quelques rappels utiles sur les différents niveaux de post-édition (light/full) et une série d’explications sur les méthodologies employées, j’ai exposé les résultats d’une étude de cas menée auprès de 25 étudiants et de 10 professionnels, à qui j’ai présenté un discours de Barack Obama traduit avec le moteur DeepL. Les objectifs étaient : 1. aller au-delà des arguments présentés habituellement par les activistes anti-NMT, 2. analyser la qualité du contenu traduit automatiquement au moyen d’exemples concrets, 3. analyser les stratégies de post-édition employées par les étudiants et les professionnels, 4. analyser les différences d’approche vis-à-vis des deux niveaux de post-édition (Light post-editing : modification du strict nécessaire en évitant toute intervention stylistique ; Full post-editing : intervention maximale pour atteindre un niveau de qualité comparable à une traduction « humaine »).
L’espace qui m’est réservé ici ne me permettant pas de reprendre en détail les exemples présentés lors de la conférence, je me limiterai à l’exposé de deux cas jugés représentatifs.
1er exemple
Le premier exemple illustre bien la fluidité de la production NMT, avec une phrase « brute » qui semble de prime abord plutôt convaincante.
Néanmoins, on se rend compte en deuxième analyse que le passage « transformer ces réalisations en un cadre durable pour le progrès humain » ne veut pas dire grand-chose.
La logique de l’exercice de post-édition voudrait donc de laisser en l’état la phrase – qui reste compréhensible – au niveau Light post-editing (LPE), mais de la modifier au niveau Full post-editing (FPE).
L’étude de cas démontre deux phénomènes intéressants : 1. 80% des professionnels ont modifié cette phrase dès le niveau LPE, ce qui laisse penser que les professionnels, probablement moins familiarisés avec l’exercice, ne peuvent résister à la tentation de corriger une imperfection, même quand ils ne sont pas invités à le faire. À noter, à titre de comparaison, qu’aucun des étudiants n’a modifié ce segment. 2. Au niveau FPE, seuls 8% des étudiants ont modifié ce segment, alors que les objectifs de la tâche l’imposaient, puisque la phrase brute n’était pas conforme aux normes de qualité que l’on pourrait attendre d’une traduction humaine. Ce constat donne à penser que les étudiants ne sont soit pas conscients de la nécessité de modifier le segment et estiment que la traduction est suffisamment bonne, soit qu’ils en ont conscience, mais ne sont pas capables d’améliorer le texte (ou ne prennent pas la peine de le faire, se contentant du niveau produit par la machine (« we are all lazy humans »).
Ces deux tendances reviennent dans l’ensemble des segments concernés par ce cas de figure, où on constate que les étudiants interviennent beaucoup moins sur le texte, aux deux niveaux de post-édition, une attitude relativement bonne dans le cas du LPE, mais qui l’est beaucoup moins dans le cas du FPE.
2e exemple
Le deuxième exemple porte sur un segment contenant une erreur de sens (contre-sens/faux sens). Comme on peut le voir dans l’illustration qui suit, DeepL a traduit la collocation « to deliver justice to the terrorist network » par « pour rendre justice au réseau terroriste ». L’erreur vient du fait que l’expression « rendre justice » en français signifie le contraire de l’original anglais, à savoir « reconnaître les mérites de », « réparer le mal qui a été fait », etc.
L’exercice de post-édition voudrait donc qu’on modifie ici cette phrase dès le niveau Light post-editing, puisqu’il s’agit d’une erreur de sens relativement grave (même si nous sommes conscients que le contexte d’énonciation ne laisse planer aucun doute sur les intentions de Barack Obama). Nous dégageons à nouveau deux phénomènes : 1. Un seul étudiant sur les 25 a pris la peine de modifier cette phrase au niveau LPE, ce qui semble très peu compte tenu des objectifs précités. 2. 20% des professionnels ont laissé cette erreur au niveau Full post-editing (et 28% d’étudiants), ce qui semble élevé compte tenu de la gravité de l’erreur et du niveau d’expertise des post-éditeurs.
Ces deux données nous amènent à penser que les deux profils de traducteurs (étudiants et professionnels) ont été « trompés » par l’apparente fluidité du texte traduit automatiquement et sont ainsi passés à côté de cette erreur grossière.
À noter également que nous avions demandé à un « groupe contrôle » de 5 traducteurs professionnels de traduire un extrait de ce texte sans recourir à la NMT et qu’aucun n’a commis d’erreur de sens sur ce passage (ni d’ailleurs sur l’ensemble de l’extrait). Il semblerait donc que l’apparente fluidité d’une production automatique induise une sorte d’excès de confiance dans le chef des traducteurs, toutes expériences confondues, et favorise donc la présence résiduelle d’erreurs de sens moins visibles.
Conclusions
L’analyse de l’ensemble du texte nous permet de dégager les tendances et constats suivants. Tout d’abord, une différence très faible entre les deux niveaux de post-édition, ce qui indiquerait que les étudiants sont potentiellement de « meilleurs » candidats pour une tâche de LPE, mais qu’ils peinent en revanche à s’élever au niveau Full post-editing.
À l’inverse, les professionnels semblent potentiellement être de « meilleurs » candidats pour une tâche de FPE, mais interviennent de façon excessive lors de l’étape Light post-editing. De façon plus générale, on note ainsi que tous les traducteurs rencontrent des difficultés à faire la distinction entre les deux niveaux de post-édition, chacun à leur échelle. Ils semblent en outre induits en erreur par le pouvoir persuasif de la NMT et se laissent plus facilement tromper par l’apparente fluidité des traductions, même lorsque la machine commet de lourdes erreurs de sens.
Nous émettons également, dans nos conclusions, l’hypothèse d’un rapport généralement inversement proportionnel entre la gravité des erreurs commises par la machine et l’effort cognitif nécessaire pour les corriger.
Ainsi, pour prendre un exemple volontairement caricatural, un contre-sens (gravité élevée) qui serait lié à l’oubli d’une négation pourra se corriger en quelques secondes en ajoutant les mots « ne… pas » à la phrase (effort faible). En revanche, des phrases bancales sur le plan stylistique (gravité faible) nécessiteront le plus souvent une reformulation de l’ensemble de la phrase (effort élevé). Cette tendance pourrait selon nous encourager les post-éditeurs à laisser les phrases en l’état lorsqu’elles sont borderline. Cette hypothèse pourrait faire l’objet d’une prochaine étude de cas.
Conclusions générales
S’agissant des conclusions générales de la présentation, j’ai souhaité insister sur les éléments suivants, en guise de takeaways. 1. Un recours exponentiel sur l’ensemble du secteur à la (post-édition de) traduction automatique, qui a véritablement le vent en poupe ; une tendance qui ne devrait pas faiblir. D’aucuns prédisent ainsi que d’ici 2030, seuls 20% de tout le volume de traduction généré à l’échelle mondiale ne passeront pas par un moteur de traduction automatique. Une prévision que je ne partage pas personnellement et qui peut sembler excessive. 2. Les post-éditeurs de toutes générations ont parfois tendance à faire aveuglément confiance à la machine, sans doute en raison de la fluidité des traductions produites. 3. Une grande confusion subsiste quant à la technologie et à son adoption par le marché, ce qui devrait inciter les traducteurs à faire preuve d’ouverture et de pédagogie vis-à-vis des utilisateurs et des « consommateurs ». Il s’agit selon moi d’un impératif pour éviter des désillusions et des grincements de dents à l’avenir, surtout chez les clients finaux, qui voient la traduction automatique comme le Saint Graal. Bien que la NMT évolue constamment – à l’heure où j’écris ces lignes, deepL ne commet déjà plus certaines des erreurs détectées lors de l’étude de cas – je reste convaincu qu’il demeurera un simple outil d’aide à la traduction, certes très utile dans certaines circonstances, mais dont l’usage doit être réservé à des utilisateurs expérimentés, formés et avertis de ses dangers.
Guillaume Deneufbourg travaille comme traducteur indépendant EN-NL>FR depuis 2002. Titulaire d’un Master en Traduction de l’Université de Mons (Belgique), il l’est aussi d’un Master spécialisé en linguistique appliquée et traductologie (ULB, Bruxelles). Guillaume enseigne la traduction à l’Université de Mons et à l’Université de Lille depuis 2010. Résident en Belgique, à quelques kilomètres de la frontière française, il est membre de l’ATA, de la SFT et de la CBTI. Plus d’infos ici.
The FLD is thrilled to share the professional accomplishments and updates of its members.
Naomi Norberg was sworn in as an expert près la cour d’appel de Bourges with a specialty in translation (more commonly known as a traductrice assermentée) in December 2018.
If you have a professional update you would like to share, please email it to the division at divisionfld@atanet.org.
By Sam Mowry
