Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces Mwando, a virtual educational assistant aimed at preserving and teaching the shiKomori language of the Comoros Islands, which encompasses four dialects. By utilizing a multi-agent architecture that integrates vector search, a knowledge graph, and web search fallback, Mwando delivers context-aware responses to user queries. Evaluation results indicate strong performance in vocabulary lookup and grammar explanations, highlighting both the system's capabilities and areas for improvement, thereby laying the groundwork for future AI applications in low-resource language education.
Mwando not only enhances the teaching of shiKomori but also sets a precedent for AI-driven educational tools in low-resource languages.
This paper presents Mwando, a virtual educational assistant designed to support the teaching and preservation of shiKomori, the language of the Comoros Islands. The system covers the four main dialectal variants (shiNgazidja, shiMwali, shiNdzuani and shiMaore) through a knowledge base constructed from phrases, proverbs, dictionaries and grammar lessons. A multi-agent architecture combining vector search, a knowledge graph and web search fallback enables accurate and context-aware responses. Evaluation on 500 queries demonstrates strong performance on vocabulary lookup and grammar explanations, while qualitative case studies illustrate both capabilities and current limitations. This work represents an initial step toward computational support for shiKomori and provides a blueprint for developing AI-powered educational tools for other low-resource languages.