More than four million young people and adults in Mexico are unable to read or write, according to data from the National Institute for Adult Education (INEA). Chiapas has the highest illiteracy rate in the country, with 11.5% of the population aged 15 and older unable to read or write—that’s roughly 11 to 12 out of every 100 people.

Additionally, the 2020 Population and Housing Census, conducted by the National Institute of Statistics and Geography (INEGI), reports that 28.2% of the population in Chiapas speaks an Indigenous language. One of the most widely spoken languages is Tsotsil, a Mayan language spoken in the Highlands of Chiapas and used by approximately 550,000 people.

Platform Aims to Strengthen Bilingual Literacy

In the highland region of Chiapas, researchers from the School of Humanities and Education (EHE) at Tec de Monterrey, in collaboration with the Chiapas Ministry of Education, have developed a technological solution to combat illiteracy in the area. The project also seeks to promote social justice and help preserve the linguistic heritage of Indigenous communities.

The initiative centers on a bilingual Tsotsil–Spanish literacy platform powered by artificial intelligence (AI) and designed to function without an internet connection, according to Alejandro Martín del Campo, researcher and leader of the project.

More than just a translation tool, the platform is intended to foster an educational process that both recognizes and honors the Indigenous language while also opening up greater opportunities through functional bilingualism with Spanish, Martín del Campo explains.

“The goal is for this platform to support a dual-literacy process. We want people in different communities to have the option to strengthen their knowledge of Tsotsil—whether by refining it, completing their learning, or even relearning it if they’ve stopped using it. At the same time, we aim for that learning to be complemented by Spanish, so they can communicate fluently in both languages,” he says.

What Can the AI System Do?

This tool will be able to:

• Transcribe speech to text in both Tsotsil and Spanish
• Translate bidirectionally between the two languages
• Convert text to speech using a native accent
• Operate offline on mobile devices or school computers

This initiative is part of the Digital Society Lab at the School of Humanities and Education (EHE), which promotes the preservation of Indigenous languages through digital humanities. The team includes educators who design learning pathways and educational materials tailored to different audiences, as well as technology specialists who develop the platform and integrate the AI components.

The project also collaborates with government partners in Chiapas and members of Tsotsil communities to review and validate the language content and support implementation in rural areas.

How Are the AI Models Trained in Tsotsil?

The system is built on a technological architecture that integrates at least four AI models based on neural networks for natural language processing. These models are trained using linguistic corpora collected directly from native Tsotsil speakers—stories, phrases, and words gathered through interviews.

Gabriela Salas, one of the lead researchers on the team, brings prior experience from a similar AI-powered Spanish–Nahuatl machine translation project that began in 2020 and was developed over the course of three years. That project achieved roughly 70% accuracy and was ultimately accepted by Google Translate. In this new initiative, she’s responsible for compiling the linguistic corpus used to train the AI models. However, she notes that “Tsotsil is more challenging than Nahuatl because it doesn’t follow Spanish logic; it’s an agglutinative language with Mayan roots, its structures are complex, and everything must be validated by native speakers before training the model.”

As for the project’s progress, it is currently about 40% complete. So far, around 17,000 phrases and 20,000 individual words have been collected. The goal is to gather at least 21,000 validated phrases to ensure that the model learns from examples that are both grammatically accurate and culturally appropriate.

How Will It Be Implemented in Communities?

Once the system is fully validated and trained, the team plans to preload the platform onto mobile devices or computers already installed in community centers and schools so it can operate offline, explains Martín del Campo. They also aim to integrate a chatbot-like feature, allowing users to interact, send messages, and receive feedback.

“The model will run in a lightweight version, preloaded onto the devices already available in the communities. Additionally, we’re exploring the possibility that, when educators visit centers with internet access, they’ll be able to download new resources or system updates, including more refined versions of the model,” he says.

Collaboration Between Tec de Monterrey and the Chiapas Government

Martín del Campo notes that the project is part of a social program called “Chiapas Puede,” through a collaboration agreement between Tec de Monterrey and the state government. This partnership enables access to communities for validating and implementing the system, while also taking advantage of the existing infrastructure in educational centers.

The team hopes that in about a year and a half, the platform will have completed pilot testing and be ready to operate in real-world bilingual literacy settings within Tsotsil communities. Through this initiative, the team also aims to show that technologies like AI can serve the common good—bridging cultures, promoting equity, and expanding access to education.

“Often, we use technology in major cities, but when it’s applied to more community-focused and social issues, that’s when its true impact becomes clear,” the researcher says.

Did you find this story interesting? Would you like to publish it? Contact our content editor to learn more: marianaleonm@tec.mx