In the heart of Colombia, two languages—Wayuunaiki and Nasa Yuwe—carry the wisdom, identity, and stories of their people. Spoken for generations by the Wayuu and Nasa communities, these languages are more than communication tools—they are cultural anchors.
But like hundreds of Indigenous languages around the world, they face a growing threat: extinction.
Now, an unlikely ally is emerging: artificial intelligence.
A New Study, A New Possibility
A recent study published in SN Computer Science (Salazar, Manrique & Pereira Nunes, 2025) explores how machine translation models—the same technology that powers tools like Google Translate—might help preserve endangered Indigenous languages.
The researchers set out to answer a bold question:
Can AI learn to understand and translate languages that have very little written data?
To test this, they focused on Spanish-to-Wayuunaiki and Spanish-to-Nasa Yuwe translation, using Transformer models—deep learning architectures known for their ability to process sequences of text efficiently.
What They Tried—and What Worked
Given the extreme scarcity of data for these Indigenous languages, the researchers:
-
Built small bilingual corpora from available sources
-
Tested simplified (shallow) Transformer models
-
Applied Transfer Learning, using Spanish-English translation models as a starting point
Their findings offer insight for anyone working on low-resource language preservation:
-
Shallow models outperformed deeper ones – When data is scarce, simpler models work better. This runs counter to the “bigger is better” approach often used in AI.
-
Transfer Learning showed mixed results – In some cases, starting with a pre-trained Spanish-English model improved results. But not always. Every language—and every dataset—needs its own strategy.
-
Data remains king – While machine learning can do a lot, it still depends on the availability of aligned bilingual texts. Without these, even the best algorithms have limits.
📘 Cite the source:
Salazar, I., Manrique, R., & Pereira Nunes, B. (2025). Machine Translation Strategies for Low‑Resource Colombian Indigenous Languages. SN Computer Science. SpringerLink
The Bridge Between Innovation and Education: Little Explorer, Big World
This kind of research aligns deeply with our mission at Luminous Photo Expeditions and our affiliated literacy initiative, Little Explorer, Big World.
Through this project, we have published and donated over 35,000 bilingual books, many in collaboration with Indigenous educators and artists. These books are now in the hands of children across the globe—including in remote communities where languages like Wayuunaiki, Sikuani, Maya, and Quechua are spoken.
Each title is more than a book—it’s a tool for connection, visibility, and pride. Our growing collection reflects a global commitment to multilingual literacy, cultural resilience, and environmental awareness.
Why This Matters
For the communities who speak Wayuunaiki and Nasa Yuwe, this research isn’t just about software—it’s about cultural survival.
Digital tools can make a difference:
-
Support bilingual education
-
Help create digital archives
-
Encourage younger generations to stay connected to their heritage
And beyond Colombia, the same lessons can be applied to other low-resource languages across Latin America, Africa, Asia, and the Pacific.