Putting People First: Community-Based Approaches to Accessible Language Technologies

The internet is a powerful tool, but for many, it remains largely out of reach. With thousands of languages spoken globally, most online content exists in just a handful, often Eurocolonial languages like English. This digital language divide is even more significant for people with disabilities, especially in regions like Asia-Pacific, where many languages are spoken and a large portion of the world’s population with disabilities resides. To take these issues center stage, at RightsCon 2025 in Taipei, Whose Knowledge? held a discussion to explore community-based approaches to making language technologies more accessible. The panel featured Khansa Maria, a disability rights activist and researcher; Rahul Bajaj, a lawyer specializing in disability rights law in India; and Asmelash Teka Hadgu, co-founder and CTO of Lesan, a company building machine translation systems for low-resource languages. It was facilitated by Maari Zwick-Maitreyi, Co-Lead of the Language Justice program at Whose Knowledge?

Khansa, speaking from her experience as a person with visual impairment, highlighted the joy of connecting with a disability community online and accessing audiobooks. However, she also shed light on a crucial barrier:

My biggest one was when participants told me when I [asked] them how do you use Urdu tech if you don't know English? And they were like, actually, that's not possible. If you don't know English, you don't get to use your tech.

This statement powerfully illustrates the inherent inaccessibility of much of the digital world for non-English speakers, even when attempting to use technology supposedly in their native language.

Rahul elaborated on his work auditing online spaces for accessibility in Hindi. He found that many platforms fall short, with a significant failure rate in technical audits. Beyond the technical, he also noted language-specific challenges in Hindi interfaces, such as literal English transliterations that lose meaning. His work emphasizes the need for developers to be receptive to user feedback.

Asmelash, through his work with Lesan, offered a contrasting, community-centered approach to developing language technologies for low-resource languages like Amharic and Tigrinya. Unlike large tech companies that might scrape existing data, Lesan actively involves community members in data sourcing and content creation. Asmelash envisions a future for African AI that is decentralized and community-driven, stating:

I hope we have different builders distributed across the continent serving their kind of language technologies, creating the best they can create.

Want to dive deeper into these crucial conversations about making the internet more just and accessible for everyone? Tune in to this episode of Whose Voices? to hear directly from our panelists as they share their invaluable insights and experiences!

This episode was produced by Milanth Gautham.

Read the transcript.

Music used in this episode is West Chamber Poem 西廂詞 by Guzhengman, CC BY 3.0, via SoundCloud.