The Barcelona Supercomputing Center (BSC) has signed a Memorandum of Understanding (MoU) with the GSMA and the telephone operators VEON and Beeline Kazakhstan in order to develop open source artificial intelligence tools through the Aina project, funded by the Government of Catalonia.

The Aina project generates linguistic resources, corpora and computer models and makes them available to everyone, so that developers of AI-based applications can work easily in Catalan. The parties to the MoU signed today will explore opportunities for cooperation focusing on the development of mechanisms and documentation for language models and AI systems and language technologies for digitally under-resourced languages, including local languages in the countries where VEON operates, building on the knowledge gained from the Aina project to provide Catalan with those resources.

This cooperation will take place within the framework of the Aina Alliance, a collaboration between the Catalan government and the GSMA to promote the presence of non-global languages in artificial intelligence models, to enable speakers of all languages to participate fully in the evolution of the digital world. The MoU with VEON and Beeline Kazakhstan is the first agreement signed within this alliance, which invites other GSMA members and mobile operators to participate in this collaborative effort to address the language gap in AI.

The Minister for Business and Labour of the Government of Catalonia, Roger Torrent i Ramió, stressed that "the Aina Alliance and this first MoU highlight the value of a project for a country that has a civil society and institutions that provide a vast wealth of data; a BSC that is synonymous with technological power and scientific excellence; an entrepreneurial and research ecosystem that is ready to innovate and adopt technology; and a Government committed to ensuring that we can live entirely in Catalan, including in the digital world". "We want to export this approach to everywhere where non-global languages are facing similar challenges," he stressed.

The signing of the MoU by the four parties took place during the GSMA's M360 Eurasia event in Baku, Azerbaijan, which was attended by the world's leaders in connectivity, for two days of discussion, debate and networking.

Addressing the language gap in AI

Only seven of the around 7,000 languages spoken worldwide have a large infrastructure in the digital world: English, Spanish, French, Mandarin, Arabic, German and Japanese. The lack of resources in other languages has created a language gap that leads to a sub-optimal user experience in AI applications, aggravating the bias in models and running the risk of broadening the digital divide in AI technologies.

Catalan is spoken by ten million people. The Aina project's mission is to create resources to guarantee the presence of Catalan in the digital world, while at the same time making those media useful for other languages with the same objective in mind. The Aina project can provide many open datasets and models.

With this agreement, VEON, which operates in countries that are home to 7% of the world's population speaking mostly digitally under-resourced languages, has made a commitment to addressing the language gap in AI by supporting the use of under-resourced languages, such as Kazakh. QazCode, the IT subsidiary of Beeline Kazakhstan, the leading operator of AI technologies in the markets where VEON operates, has launched an AI module in Kazakh that supports customer services in the operator's digital applications.