Notes from our first field deployment of Bridge for messaging apps
by An Xiao Mina, Director of Product
As a frequent traveller, I find that one of the biggest challenges is communicating basic things with other people. Sometimes all I want to ask is “How much does this cost?” or “I am looking for a USB stick to purchase.” Sometimes Google Translate can handle this but often the challenge is: how can I be sure? For some languages, like Spanish, machine translation to English can be very effective. But for others, like Arabic or Farsi, machine translation quality is inconsistent at best, and when the content is conversational or informal, the result is often unintelligible or inaccurate.
Refugees entering a new country or linguistic sphere face these challenges each day. When the Meedan team visited a refugee camp in Serbia, we found that basic questions, like “Does this food contain pork?” or “What time does children’s clothing distribution begin?” become more complex with language barriers between refugees and aid workers.
At Meedan, we’ve designed Bridge to address a lot of these challenges. Bridge empowers human translators to translate short bits of text. Because we value accuracy and speed, we’ve tried to optimize the user experience to enable translation as quickly as possible, with simple tools like machine translation and dictionaries, to ease that process along.
We’re pleased to announce that we have recently begun testing Bridge’s capacities for translating short messages on messaging apps, with a special focus on how that can improve communications between aid workers and refugees. It was important to us that users not be required to download a new app, and we already saw that WhatsApp and Viber were the most popular amongst refugees. Of these apps, the latter is the only one with a useable API.
For the purposes of the closed pilot, we are testing use cases for this process and the effectiveness of how this tool works with mobile messaging apps, and we are not handling private or sensitive messages at this time. In the longer term, we plan to include broader messaging app support, including apps that support end-to-end encryption, photo sharing and other key features.
It’s a simple process:
- The requester types a short message on Viber with our Bridge Bot
- Bridge routes the request to a human translator
- A human translator makes a quick translation and, if necessary, an editor reviews it
- The translation returns to the requester, with both text and an image card, which can be saved for future reference
Here’s what that looks like in practice:
Sending a Request
In Viber, type a short message that you would like to have translated. In this case, that message is “What is that?” in English. This particular bot is our English — Arabic bot, so the request gets routed to English-Arabic translators, with a short acknowledgement of receipt.
Receiving a Response
After a brief period, a human-generated translation arrives on Viber. It takes the form of text, as well as an image card. The card can be saved for future reference and held up to a reader as a simple way to share a message in a language you don’t speak.
This card format in particular is designed for the refugee camp context. It’s based on studies we did in our previous field research on how aid workers might be able to communicate basic announcements and messages as they make their rounds at a camp. It’s inspired as well by the Translation Cards tool built by Thoughtworks and Google.org, a critical tool for repetitious words and phrases in multiple languages.
The use case for Bridge right now is for humanitarian workers — in partnership with MercyCorps and with support from the Cisco Foundation, we spent an extensive amount of time interviewing both refugees and aid workers in Serbia about the translation challenges they face and how they benefit from simple, quick text-based translations on a popular messaging app like Viber.
We’re now in the testing and implementation phase. In Berlin, colleague Abir Kopty and I have been working with The Space Berlin and other community members to do focused testing. What works with mobile messaging app translation? What are the most commonly-requested phrases? How can we streamline this process for someone who might be very busy, especially in a refugee camps?
And looking forward, we hope to understand different use cases and technical features. We’re starting the pilot with Viber but starting to look at other platforms widely used amongst refugees that have API support and useful features like end-to-end encryption and image sharing. We’re considering, for instance, audio input and audio output, and how that might improve the process. We’re also curious what other applications you see for a tool like this.
Do you want to help translate? Do you want to help test this app? Want to partner with us? Please get in touch at firstname.lastname@example.org, and we can talk more.