The second day of the Language Diversity Conference 2026 continued today, Saturday, 3 October, in Accra with another full program of discussions, workshops, and practical sessions focused on the future of languages in the digital world.

It’s the first Saturday of the month, which in Ghana means national clean-up day: shops close, and people spend the morning tidying up their neighbourhoods. If the Ghanaians in the room had been at home, they’d have been out cleaning too.

There was a quick word about the food, too. So far we’d been served peppers in red, green, and black, the colours of the Ghanaian flag. The yellow ones? Still to come. Different shapes and different colours, a bit like everyone in the room.

Then it was down to business and to a question that has been hanging over this conference since day one: what does AI mean for the languages people here are working so hard to keep alive?

“We give our words”

Sucheta Salgaonkar, Lead Product Manager for AI and Machine Learning at the Wikimedia Foundation, started with something personal: her grandparents’ letters. She grew up between languages, but after her family moved from India to the US, her parents switched to English to fit in. Now in her thirties and learning Marathi, she reads the letters her grandparents sent her father by scanning them with Google Translate’s camera. She can do that only because Google Translate handles Marathi well, likely thanks to Marathi Wikipedia and its editors.

That was her point. For some languages, Wikipedia may be the only place they exist as digital text, so a lot of what AI knows about them comes from Wikimedians. So what do those languages get back? “The honest answer is right now, not enough.”

Sucheta Salgaonkar, Lead Product Manager for AI and Machine Learning at the Wikimedia Foundation, speaking at the Language Diversity Conference 2026, Day 2.

She was clear that nobody has to use AI. But for communities that want to try, she offered a ladder of four ways in, starting with tools that work in any language:

models that judge articles by their structure rather than their text

machine translation through content translation

editing suggestions such as Tone Check

Open models the Foundation is hosting through Project Hummingbird

Her ask was to pick one rung and try one small experiment this month.

“We don’t have to wait for external commercial services to prioritize us before we prioritize ourselves.”

The Q&A went straight to the hard stuff, including how small communities can say no to AI-generated articles without putting off new editors. Her answer was to build guardrails into the editing experience and to leave decisions about AI to each community.

Bringing more women into minoritized language Wikipedias

One morning session focused on the gender and language gap.

“Bridging the Gender and Language Gap: Scaling African Women’s Participation in Minoritized Language Wikipedias,” explored efforts to increase women’s participation and contributions across African language Wikimedia projects.

The discussion drew on initiatives involving languages including Setswana, Kiswahili, Luganda, Twi, Kinyarwanda, Runyankole, and Malagasy. It also looked at challenges such as language accessibility, limited local-language content, and the cost of internet access.

The morning also included a practical workshop on Wikidata Lexicographical Data, using Dagbani and Igala as examples. Participants were introduced to lexemes and learned how structured vocabulary data can help document and preserve underrepresented languages.

Another session explored the possibility of moving from Wikipedia to mother-tongue AI, using the Igala language as a case study. The discussion looked at how native speakers can play a direct role in guiding and improving AI systems for languages with limited digital resources.

Practical tools for language communities

Day Two also provided several hands-on learning opportunities.

A session titled “From Draft to Published: An AI-Assisted Translation Workflow for Wikipedia” introduced participants to ways of combining translation tools with Wikipedia’s own tools. The session also highlighted the importance of checking AI-generated translations, especially for low-resource and Indigenous languages.

Participants also explored “My Quechua is not on Wikipedia,” which examined representation, content, and community ownership within the Quechua Wikipedia. The session raised broader questions about how Indigenous language Wikipedias can better reflect the different varieties and experiences of their speakers.

The session also brought together leaders from different language communities, including Igbo, Dagbani, and Fon, to share experiences and lessons that could support the development of the Baoulé Wikipedia project.

The program also featured a presentation on the Dagbani Wikimedians User Group, highlighting programs, achievements, and challenges from 2025–2026.

Audience at the Language Diversity Conference 2026, Day 2.

Dictionaries, accessibility, and inclusion

After lunch, the program continued with a series of shorter presentations and practical sessions.

One presentation, “ Anyone can Make a Dictionary, ” looked at simple ways communities can begin creating dictionaries without treating dictionary-making as something reserved only for specialists.

The program also explored the relationship between Wikimedia projects and language documentation and revitalization, using Kihunde as an example.

Accessibility was another important part of the afternoon. Wiki4Inclusion highlighted efforts to make Wikimedia more accessible to underrepresented communities, including persons with disabilities.

A separate session on Hausa Sign Language examined the need for better digital resources for Deaf communities and explored developing a digital corpus to support future language technology.

The conference also looked beyond language documentation into access to important information. WikiMed Africa presented work on making medical information more accessible in African languages through Wikimedia-based platforms and community contributions.

Building digital infrastructure for languages

One notable afternoon presentation focused on the Dagbanli Dictionary, an open-source, offline-friendly digital language resource.

The project uses structured data from Wikidata Lexemes, along with audio recordings, to create a digital dictionary that can also work offline. The presentation showed how language communities can build their own digital infrastructure rather than relying only on conventional bilingual word lists.

Another session, “Breaking the Metadata Barrier,” looked at the importance of translating Wikidata properties, labels, and descriptions. The workshop demonstrated how language communities can make the less visible parts of the digital knowledge ecosystem more accessible in their own languages.

The workshop also covered community engagement, language documentation, and the experience of the Cebuano Wikipedia, showing how human relationships and community participation remain important even as technology continues to develop.

Looking at language technology with care

As the afternoon progressed, several sessions returned to the relationship between language and technology.

The presentation “The Power of Speech-to-Text Technology in Preserving Minoritized Languages” examined how speech-to-text tools can help document oral histories, stories, songs, and local knowledge, while also recognizing the limitations of current technology for smaller languages.

Another presentation examined the challenges of AI translation for African languages and the possibility of fluent-looking translations still carrying inaccurate meanings. The discussion highlighted the importance of language data and the active involvement of language communities when developing and using these technologies.

The day also featured a session on building partnerships for sustainable language documentation and revitalization, drawing on the Torwali experience in northern Pakistan. The presentation emphasized community ownership, long-term partnerships, and the importance of keeping languages in active use.

Speaker with Sign Language Interpretation support at the Language Diversity Conference 2026, Day 2.

More than presentations

Day Two was not only about listening to speakers.

Participants had opportunities to take part in workshops, ask questions, share experiences, and work directly with tools. There were also incubator office hours for people seeking help with developing new language projects.

Later in the afternoon, participants had the opportunity to take part in sightseeing around Accra, followed by the evening dinner listed in the program at Fiesta Royale.

The range of sessions throughout the day showed that strengthening a language online requires more than creating Wikipedia articles. It can involve dictionaries, audio recordings, Wikidata, translation, accessibility, offline technology, AI, community organizing, education, and partnerships.

Continuing the conversation

Day Two brought together experiences from Africa, Asia, Latin America, and other parts of the world. While the languages and communities represented were different, many of the conversations shared a common concern: how can communities take an active role in shaping the digital future of their languages?

From Igala and Quechua to Dagbani, Dameli, Torwali, Javanese, and many other languages, the sessions showed different approaches to documentation, revitalization, and digital participation.

The second day therefore continued the conversation started on Friday while giving participants more practical tools and ideas to take back to their own communities.

With one more day of the Language Diversity Conference remaining, conversations in Accra continue to connect people, ideas, and experiences around one shared goal: strengthening language communities in the open knowledge movement.

Exploring Accra beyond the conference

Day Two also gave participants the opportunity to step away from the conference room and explore some of Accra’s important historical landmarks.

We visited Osu Castle, then proceeded to Black Star Square, before ending the tour at the Kwame Nkrumah Memorial Park and Mausoleum. It was a good opportunity to learn more about Ghana’s history, heritage, and journey while connecting with fellow participants outside the conference sessions.

It was a simple but memorable part of the conference experience.