Nieman Foundation at Harvard
If Brexit drove home any point, it’s that the fates of Europe’s countries are deeply intertwined. Decisions made in Brussels — along with happenings in capitals across the continent — affect millions of people in dozens of countries. But for news organizations with limited resources though, covering all aspects of European policymaking — across border, across languages — is a major challenge.
In an attempt to provide more thorough coverage, outlets across the continent are banding together to share coverage — but with 24 official languages spoken in the European Union, reusing someone else’s reporting often means that publishers need to translate stories and adapt them to an audience in a different country. And translating content presents its own challenges; the process of adapting stories from country to country can be costly and time consuming.
“We knew that [translation] was going to be a problem for the start,” said Javier Moreno Barber, director of the Leading European Newspaper Alliance, a partnership between seven news organizations in six countries that publish in four languages. “Especially because of the money. It’s expensive.”
As a result, a handful of different news organizations and partnerships are trying different approaches to translation.
The seven LENA papers — Germany’s Die Welt, Spain’s El País, Italy’s La Repubblica, France’s Le Figaro, Belgium’s Le Soir, and Switzerland’s Tages-Anzeiger and Tribune de Genève — have full rights to repurpose any work the other outlets publish, but each individual paper is responsible for translating the stories or multimedia pieces they want to use. For the first time this fall, LENA will try a centralized translation process for its coverage of the U.S. election.
The policy news site EurActiv, meanwhile, employs five full-time translators in Brussels, Paris, and Berlin. It also has a network of 12 franchised sites in other capitals around Europe and it has worked with other outlets such as The Guardian and Der Speigel. EurActiv is also experimenting with machine translations that it hopes will make its processes speedier and more efficient, despite lower quality.
“More and more media are translating some of their content into other languages because they want to reach a broader audience,” Gian-Paolo Accardo, VoxEurop’s editor said. “What we’ve noticed — and this is one of the reasons why we thought about setting up this network and being the center of it — is that some media don’t really know or can’t figure out what is interesting for an audience they don’t know. Most of the media know their audience when they are local or small media — they know their audience quite well — but once they want to reach an audience that is outside their market, we’ve noticed that they often translate content that this audience is not interested in or they don’t localize it enough — that, for example, some people’s names are not detailed, political leaders are not specified. This is something that needs an expertise.”
VoxEurop is developing the network, called the European Independent Media Network, as a European Cooperative Society. The site emerged out of PressEurop, the European Commission-funded site in 2013, and it hopes that transforming into a larger network will allow it to attract more funding and have a larger reach.
VoxEurop says it now receives 1.25 million unique visitors annually. It publishes in 10 different languages, but English, French, and German are the most popular languages.
News outlets that want to participate in EIMN’s content sharing and translation services will have to buy a share of the co-op, pay an annual fee, and then an additional cost for the translations they wish to use.
“The purpose is to provide them with a platform for content sharing, translating, localizing, and also making the network live through an internal newsletter,” Accardo said. “Meaning a Slovenian media can be aware of what its French partner is publishing that might be interesting for him to reuse and translate it to Slovenian.”
The network will utilize VoxEurop’s existing translation infrastructure. The site works with independent translators; they’re volunteers and aren’t paid.
However, one of the site’s main revenue-generating tactics is offering paid translation services aimed at other media outlets. VoxEurop will offer the translators it works with for editorial content first dibs on those paid jobs.
“They are quite devoted,” Accardo said of the translators. “One of the reasons VoxEurop is still running, and that it has been running for now more than two years, is that the people who are volunteering for it have a stake in it. A small financial stake, being the opportunity to get paid jobs through it, but also they think it’s a project that deserves to go on, and we — the management and the five people who are part of the core team — are fully committed to finding proper assignments for third parties and other sources of financing.”
Like VoxEurop, the Leading European Newspaper Alliance is also establishing its own centralized translation processes. This fall, for the first time, LENA will experiment with a unified translation service for its coverage of the U.S. election.
The seven participating papers will share reporting on the presidential race — but on a day-to-day basis they already have access to everything each other paper publishes. Some papers use more LENA content than others; Moreno said there’s no overall count of how many stories have been shared or translated, but he estimated that some publications use up to seven or so LENA stories per week.
There has been some unified collaboration between LENA papers. Last month, for instance, El País, Le Figaro, and La Repubblica each published a joint interview with Madrid mayor Manuela Carmena in their respective languages.
Most of the translation, however, has been on an ad-hoc basis and each LENA news organization pays for its own translations and uses its own processes, which differ from paper to paper and country to country. Still, since there’s some language overlap among the publishers, there’s often no need for translations — and some of the viral-type videos that the alliance shares work no matter the language they’re in.
Each LENA paper pays a set annual fee to join the alliance. (Moreno would only say that “it’s not a huge budget.”)
The alliance’s U.S. election coverage will be the first time that it pays for translation; Moreno said LENA has contracted with a French firm. Each paper will have its own correspondents in the States reporting on the race, but LENA also will pay for a handful of reporters to write stories exclusively for the alliance that will be translated into all four languages and available for the papers to use.
Moreno stressed that this is just a short-term experiment, but if it goes well and there’s interest from the papers to continue with a unified approach, he said he could imagine it could continue.
“With my budget, I cannot really pay for translating everything we’re running in all seven newspapers,” he said. “If the newsrooms think this is a very good approach and they think that this is something which helps the alliance, obviously I will have to go to the publishers themselves and say, ‘This is what the newsrooms are saying. We’ve tried that with the U.S. election. It worked. We’re very happy, but obviously we need more money to do that.'”
“That will be a second step,” Moreno continued. “For me, it doesn’t make any sense at all to start having a system…to translate things before we even had tried to see if there was interest or if it would work or not.
With its team of full-time translators, EurActiv has been honing its translation prowess since the site debuted in 1999. The site translated 2,750 stories last year, but it is working on making the translation process more efficient. The network averages about 794,000 unique visitors a month.
One way it’s working to do that is through machine translations. EurActiv is working with a Latvian company, Tilde, that is ingesting tens of thousands of EurActiv’s stories and human translations to teach the computer about the types of stories and language the site uses.
“It helps the machine learn the kind of things we write about and then produces higher quality suggestions when it does a machine translation of an article,” EurActiv translator Samuel White said.
EurActiv said it was in the very early stages of developing the machine translation processes, but David Mekkaoui, director of the EurActiv foundation, said the site hopes the new technology will make the translation process three times faster.
Still, anyone who has ever used Google Translate knows that machine translations can be buggy, and White cautioned that it will take some time for the machine translations to work seamlessly. He pointed out that much larger organizations, such as the United Nations and European Commission, have been trying machine translations for years and are still working out the kinks.
The computer has a lot to learn; there are phrases that work in one language that aren’t ideal when translated. For instance, there’s lots of repetition in English — it’s okay to write, “Obama said this,” “Obama said that,” etc. — but in French, that wouldn’t work stylistically, White said. Similarly, while English is a very linear language, the same phrase in German can be arranged grammatically three or four different ways. That’s a “huge challenge for a computer,” White said.
“The challenge for a small-scale operation is actually building up that memory, that kind of translation memory, particularly in the languages we translate less frequently, like Spanish or Italian,” White said. “Based on our current way of working, it’s going to take quite a long time to build up the machine to a point where it’s really reliable.”
Aside from just machine translation, EurActiv is experimenting with tagging and semantic web tools to create a more streamlined translation process and build a platform to facilitate translations.
“Picture yourself as a journalist writing an article about translation, and another colleague in the media that’s a partner has also written an article about translation, but he wrote it in Spanish,” Mekkaoui explained. “The machine would be able to track down that article, understand the topic of it, and once you put in your platform that you’ll be talking about translation, it’ll tell you: Here’s another article about translation, and this is a rough translation of this other article, purely machine translated, so you can go through it, read it, and see how relevant it is.”
The reporter, Mekkaoui said, could then decide to use the rough machine translation as background information for her own story. Or, if they decide that the foreign language version is sufficient, they can send the story to a translator who’d adapt it.
With journalists and translators working together, the aim is to improve the localization process, to make sure that there is enough context and description for stories to work in multiple languages. EurActiv calls the more involved translation process post-editing.
“That translator will see the wild translation of the article and then compare it with the initial source,” Mekkaoui said. Then, after more work, “we believe there is one more step that is needed: localization. That article will then go to an editor or journalist, and that editor or journalist will be able to make sure that the content is relevant for his audience.”
Despite the costs and logistical headaches, all of these efforts — along with the work VoxEurop, LENA, and others are undertaking — are undertaken with the goal of getting as much journalism in front of as many people as possible.
“What we try to achieve is to improve the amount of content among different partners,” Mekkaoui said.
