How Do You Say…? Notes and First Impressions of Translation Challenges in a Refugee Camp in Serbia
written by Abir Kopty and An Xiao Mina
A refugee camp is its own world of complex logistics: registering newcomers, distributing food and clothing, providing medical support, delivering education programs and other services. One of the compounding factors is language: aid workers and refugees often come from very different contexts, and they often do not share a common language.
Recently, Meedan started working with Mercy Corps to test out the possibilities of technology in helping overcome language barriers. We knew that understanding the context of the refugee camp — its ins and outs, the relationships between refugees and aid workers, and the daily language barriers — would be critical to success. We spent some time in a refugee camp to test how Bridge, a tool that helps humans translate conversational content online, could best serve best both the refugee community and the aid workers trying to help them.
Bridge started with social media, allowing different translation communities and groups to translate posts and then share those translations back to social media. For instance, the Global Voices Lingua community used Bridge to translate social media for RightsCon, the world’s largest human rights and technology company, to help Spanish and Arabic speakers; and the Translators Without Borders community used Bridge to translate social media for the Out of Eden Walk project, and those translations are then embedded on the project’s site. It’s an ideal platform for translations that require an understanding of nuance and cultural context.
In November 2016, we traveled to the town of Presevo in south Serbia and spent a week in its relatively small refugee camp (about 500 refugees), to understand the challenges in the field that both refugees and aid workers face when trying to communicate with each other. The research sought to understand how we should focus the work of Bridge, to facilitate more effective communications across languages. The relatively small size of the camp makes for ideal conditions for a pilot, reducing the number of variables and challenges for implementation in a limited trial setting.
Though Bridge was first designed for social media translation, we recently started adapting it to handle content from messaging apps, including audio, text and images. Messaging apps matter — more than 80% of Syrians in Europe use smartphones, and the majority of refugees we spoke with rely on tools like Viber and WhatsApp to stay in contact with family and coordinate basic logistics on their journey. These apps are already used in many camps, and so we hoped Bridge could piggyback onto these tools’ APIs (if they exist) to provide better translation support for both digital and in person conversations.
These were our goals:
- Understand the need for live translation in the field when using chat apps like WhatsApp and Viber
- Understand the need for pre-prepared translations in the field developed for Translation Cards
- Document and observe daily life in the camp, with a specific focus on language and access
- Build a working relationship with refugees and aid workers, while identifying potential refugees to participate in the pilot as translators
Stage 1: Focus Groups and Interviews
Over the course of the week, we held three mixed-gender focus group meetings with refugees from Syria, Iraq and Pakistan. We worked with these groups on identifying use cases for translation needs, testing translation in certain scenarios and identifying solutions. We also conducted informal interviews with several aid workers in the camp, and we participated in observing certain activities and assessing the translation needs.
Refugee camps are busy, linguistically diverse places. At Presevo Camp, translation is constantly needed. The camp can be a place of routine but also a place full of vibrant dynamics and interactions. Language barriers add extra burdens to daily life at the camp, from the simple (clothing and food distribution) to the complex (medical and psychological care). Most, but not all, aid workers speak English and another European language.
The majority of refugees at the camp speak Farsi, Pashto and Urdu, languages spoken in Iran, Afghanistan and Pakistan. A minority speak Arabic and Kurdish, languages spoken in Syria and Iraq (among other places). A very small minority of refugees speak English, and very few speak other languages beyond their native tongue.
Professional interpreters are extremely scarce, though some organizations employ interpreters, who usually translate from/to English or from/to Serbian or Albanian. Refugees who have capacity with English and their native tongue regularly serve as important bridge figures between refugees and aid workers. An even smaller minority of particularly talented individuals have capacities in multiple languages. However, both availability and capacity to serve all translation needs in the camp are very limited.
Refugees recounted to us several scenarios in which translation could overcome day-to-day language barriers, from shopping, navigation, finding a restaurant or cafe, talking about their needs or hobbies with the aid workers, to medical needs, distribution cases and more. Aid workers validated many of these refugee translation needs too, but they also talked about their own translation needs, including the need for basic conversation, disseminating important logistics and health information, sharing instructions during workshops, and answering questions from refugees.
Stage 2: Testing Live Translation
The basic idea is simple: whenever a refugee or aid worker needs a translation, they can make a request via a messaging app. Bridge takes that request, routes it to an available translator, and then routes back the vetted translation. We have confidence in this workflow when it comes to social media, but we’d never tried it for messaging app.
At Meedan, we ground our design in people-focused research. Before even beginning work on this new aspect of Bridge, we wanted to ground it in some real world use cases. So we developed a simulation: using WhatsApp, refugees and aid workers could request translations. Instead of routing them through Bridge, we just sent them directly to Abir (co-author of this post and co-researcher), who translated them quickly.
Our top level insights? Translation is necessary, and in all sorts of formats: text, audio and image. In most cases, participants preferred to send audio requests rather than writing phrases, because this could be the fastest and easiest option for them. They also suggested that in some cases they would want to send an image (particularly of road signs) and get its content translated.
We learned too that some translations might need a timely turnaround — within minutes — while others can be prepared ahead of time and require a one day turnaround. And here’s how they might look in WhatsApp:
We tried out a few scenarios:
At the Grocery Store
Refugees often need to go to a grocery store to pick up daily supplies. Because refugees could not leave the camp, we had to try out these scenarios in the camp. One of them was a simulation exercise for a grocery shopping scenario. The refugees interacted with a shopkeeper (that’s An, the other author of this post and co-researcher) while a translator (Abir) was available on WhatsApp sitting in another room. Here’s how it worked:
- They were shopping for groceries with An, who doesn’t speak Arabic.
- When they needed help, they sent translation requests to WhatsApp using the audio feature.
- They often asked in Arabic “how do you say…. In English”.
- The translation came back as an audio file or text with “one kilo of tomatos” or “strawberry” or “I need 2 chocolate milks,” etc.
- When An needed to communicate something back in Arabic, she follow the same process, speaking in English and getting an audio response back in Arabic that she could play back.
Having a Conversation
We also observed that in some cases, where the refugee or the aid worker speaks a certain language (say Arabic or English) at a good level, our tool might be vital when they are stuck with a word or two. So we tried having a long, complicated conversation in English, with two refugees who were fluent in English. Again, they were sitting in one room and a translator was in another receiving and processing their translation requests. Our testing suggested this technology might be of a great help when they get stuck and need a word or two in English.
Stage 3: Testing Translation for Routine Activities
There are many routines at the camp: distributing clothes and food, offering medical services, and other activities. With hundreds of people, conditions might be tense and hectic. In these cases, on-site translation might not be the best way to cope with translation needs. The solution we suggest is pre-prepared translations. This can happen via Translation Cards, an app developed by Thoughtworks and Google.org, or by using messaging apps. We have tried both.
Imagine a line of 300 people waiting for clothing distribution. In this case, aid workers often repeat the same basic information for each person, in languages like Arabic, Farsi, Pashto and Urdu. These sentences include for example “we have no more shoes”, “we only have jackets for babies” or “we do not offer this or that service here, please go to x or y offices”. Translation for such phrases can be pre-prepared and then used to response to refugees’ expected inquiries.
We used two methods. The first was preparing cards for Translation Cards. After interviewing aid workers about commonly used words and phrases, we prepared the translation of common English phrases into Arabic, Urdu, Farsi and Pashto, in two formats, voice and text. Aid workers could then simply tap on a button to play a card’s audio. There was a logistical challenge — being able to hear the audio in an open air environment — so we brought in a Bluetooth speaker to help amplify it. Overall, workers found this greatly helped with their communications.
The second was using simple messaging apps to handle requests of common phrases, and while preparing we realized the need for the translation in three forms: audio, text, and image. Being able to take in this kind of content and produce a translation package — a file with text, audio and a mobile-ready text image — allows for maximal flexibility for users. Text can be copied and pasted into another app; audio can be played aloud (so long as there are strong speakers!) for people with limited literacy levels; and a mobile-ready text image can be shown at full screen on a phone, or printed out and pasted on a door.
Moving Forward
There are many forms of translation at a camp: live interpretation and phone based interpretation are often the most familiar and best suited for complex and nuance-rich conversations, but they also require time and resources that not all camps have access to. At the same time, not all translation needs are high-touch; some are repetitious announcements, and some just need to convey basic information. As well, a number of refugees and aid workers we met have a basic capacity with each other’s mother tongues but need help with some phrases here and there. With the introduction of messaging apps and custom tools like Translation Cards, some (but not all) translation needs can be handled in a more distributed way, relying on networks of volunteer translators with an ability to contribute their skills.
These important findings mean that we should adapt our technology to these needs, which we identified in concert with both refugees and aid workers through regular feedback and discussion. Now that we have formed some relationships with people at the camp, we are hoping to be able to handle remote translation requests through the help of our translation community. We are also working on translating social media information on refugee issues into languages useful for the refugees or the aid workers. In the long term, our overall hope for our technology is to contribute to global efforts among civil society to serve the refugee and aid community, by enabling them to communicate and convey their needs/service without the barrier of language.
