AI and Education
Building AI-powered learning tools for low-resource languages.
Overview
Zoorna’s AI and Education initiative develops intelligent educational tools tailored to the unique challenges of teaching and learning underrepresented languages. Our pilot platforms support Persian and Armenian, with future expansions planned for Kurdish, Tajik, and other minority language varieties.
In a world where educational technology is overwhelmingly designed for high-resource languages, this project aims to bridge the digital divide by creating AI tutors, vocabulary tools, and interactive reading assistants that are both linguistically informed and culturally aware.
Methodology
Our approach integrates:
- Prompt-based Large Language Models (LLMs) to create adaptive question-answering and summarization tools
- Narrative and syntactic analysis for curriculum-aligned reading comprehension support
- Paraphrasing and vocabulary scaffolding tuned to second-language and heritage learner needs
- Custom annotation tasks to evaluate model performance on complex linguistic phenomena (e.g., aspect, modality, politeness, narrative structure)
- Iterative prototyping with student and educator feedback loops
- Multimodal interface design to support multilingual input/output and explainability
We're also developing automated pipelines for expanding lexical resources, glossaries, and test batteries.
Preliminary Results
- GPT-based reading tutors can accurately summarize and generate comprehension questions for Persian texts at intermediate-to-advanced levels
- Vocabulary support modules improve learner confidence, especially among heritage speakers
- Prompt structure and pedagogy-aware tuning significantly affect outcome quality
- Students report high engagement with culturally relevant examples and intuitive interfaces
- In Armenian, verbal morphology and dialect variation pose specific challenges, which we are addressing via fine-grained morphological analysis
Use Case
Our AI tutor framework can be deployed in:
- Language classrooms teaching Persian, Armenian, Kurdish, and other low-resource languages
- Heritage language education programs seeking tech-enabled curriculum
- University courses focused on Middle Eastern or Caucasian studies
- EdTech products looking to expand into underserved linguistic markets
- NGO or diaspora community workshops promoting native language maintenance and digital literacy
- Individual learners interested in advancing their reading comprehension and vocabulary skills
Long term, we aim to release modules and educator-facing tools that plug into existing learning management systems (LMS).
Team
Project Lead: Dr. Karine Megerdoomian
Team: E. Garcia, Andrew Megerdoomian
Pilot Institutions: UC San Diego, U of Chicago, U of Toronto (planned)
Latest publication or presentation
(if available)