I am Alp Öktem. I earned my PhD in 2019 in the TALN group at UPF. During the beginning of the COVID pandemic, I designed and led the project SuportMuTu, which aimed to extend the reach of information and neighborhood solidarity channels, with artificial intelligence-powered collaborative translation.

During my studies I co-founded the non-profit knowledge cooperative Col·lectivaT where I currently work as a computational linguist. My co-founders and I are all from migrant origins. During our migratory paths, we have developed awareness of the role of language in society in our own ways. I realized through my personal and professional experiences that a language can open doors but it can also close some. Which language we speak often determines if we have access to information or not and to what extent our voice matters in society.

"The project SuportMuTu aimed to extend the reach of information and neighborhood solidarity channels, with artificial intelligence-powered collaborative translation."

Whose language is excluded from public spaces is eventually excluded from technology and that’s where my work lies. I create linguistic resources and tools for marginalized and minority languages so that they are not excluded from the digital domain. During my post-doctoral career, I have had the chance to work with many languages including Catalan, Kurdish, Syrian Arabic, Tigrinya, Amharic, Hausa and Swahili.

Languages of power

The start of the COVID-pandemic was disorienting for most of us. We all saw that the threat wasn’t only a deadly virus, but also lack of information, misinformation, restricting public measures, economic uncertainty and loss of proximity with our neighbors and loved ones. I remember the intense period at the beginning where I was calling everyone I knew, trying to predict what was going to happen and establish some facts to keep myself sane and safe. I was privileged enough to be able to inform myself in four languages: Turkish, to get informed on what was going on in my country of origin, English for following global news and finally Spanish and Catalan to follow local information and measures.

The start of the COVID-pandemic also meant lack of information, misinformation, restricting public measures, economic uncertainty and loss of proximity with our neighbors and loved ones.

What gave me a sense of hope in this dark period was seeing how activist groups of Barcelona were taking their initiatives online to grow solidarity networks in their neighborhoods. I remember being invited to a Telegram channel in my neighborhood, Vallcarca, and then to a WhatsApp group on my old street in Gràcia. I later discovered similar ones from other neighborhoods such as El Raval, Sant Antoni and Sants. Just like official health channels, these groups were making regular announcements on how to keep safe and where to find help. Also, they were being used as noticeboards connecting those in need to those who can help, be it for food or domestic appliances. People were supporting each other in creative ways, adapting to the harsh measures of confinement.

Who controls the public conversation?

What I was seeing wasn’t only hopeful to me but also quite innovative. Although, through the eyes of a migrant, it didn’t take me long to notice that these efforts were not inclusive enough. The main language in these channels, both official and neighborhood ones, were exclusively Catalan. This meant that those people that do not control the language very well, like many citizens of non-European and non-American origin, were practically left out of the conversation.

"Those people that do not control the language very well, like many citizens of non-European and non-American origin, were practically left out of the conversation."

The monolingual structure of public conversation and spaces has a clear marginalizing effect to those from diverse origins. I was witnessing this very sociolinguistic exclusion replicating itself in this newly formed virtual space and thought that I could do something about it.

Could artificial intelligence counter exclusion?

During my free time, I came up with some simple software that could mirror Telegram channels to serve in other languages without any need for human intervention. This solution used machine translation, a technology that I have been working on since my doctoral studies. However, I saw that translations made by MT are not 100% reliable, especially in low-resourced languages.

"By adding speakers of the translation languages in the loop, the result of machine translation could be more accurate and reliable."

Seeing its limitations with my colleagues in Col·lectivaT, we decided to extend the technology to involve speakers of the translation languages. By adding them in the loop, the result of machine translation could be verified before they were sent out to the translated channels.

We optimized this process by building a bot that listens to messages being sent to original channels, automatically translates them and then sends them to linguists for approval. If a message needed any corrections, they could correct it and then the bot would take care of its delivery to the mirrored translated channels.

This ensured 100% accurate translations without going through a fully-fledged translation process. And to make it further accessible, the redirected messages were accompanied with a spoken version delivered by a synthetic voice, so that people with reading disabilities could follow as well.

An example that shows the path of a message published in the official channel of Catalonia’s Health Department, reaching the Arabic speaking public.

SuportMuTu

In a short time, our network started serving the Catalan health department channel and various neighborhood channels in Arabic, Urdu and Chinese, some of the most common non-latin migrant languages of Barcelona.

We decided to keep the name as used by the neighborhood support groups: SuportMuTu (mutual aid). This is a term coined by anarchist philosopher Peter Kropotkin in his essays arguing that evolution is driven by cooperation and not competition. Communities organized around mutual aid prioritize meeting each other’s needs in an egalitarian and voluntary fashion. Our work was only the addition of technology to this picture, hence the uppercased M and T, which stands for machine translation.

"Our initiative was highly welcomed by the neighborhood groups we served as it attracted people from diverse origins to their channels."

Our initiative was highly welcomed by the neighborhood groups we served as it attracted people from diverse origins to their channels. We were able to sustain the project during the peak period of the pandemic thanks to the crowdsourced Cooperative Social and Health Emergency Fund (Fons Cooperatiu per l’Emergència Social i Sanitària) as well as other projects that belong to the Catalan social and solidarity economy.

As a blueprint for digital inclusion

SuportMuTu has been a unique way of responding in a very unique time demanded by the pandemic. For many, the introduction of confinement measures has been a radical change to the digital realm –be it for working, studying, organizing or getting informed.

SuportMuTu showed that this change could be made in a way that doesn’t make language a barrier and even serves to create bonds in diverse sections of society. We believe that it could be a blueprint for increasing citizen inclusivity in areas like education, health, governance. Further results of the project can be consulted in this blogpost.