An AI-Powered, Multilingual Pedagogical Support Framework for Integrated Science Teacher Education in Resource-Constrained Nigerian Contexts

Ismail Ado Funtua; Mahmud Muhammad Yahaya; Aliyu Dahiru Karofi

doi:10.58709/niujed.v12i3.2715

Ismail Ado Funtua Isa Kaita College of Education, Dutsin-Ma, Katsina State, Nigeria
Mahmud Muhammad Yahaya Isa Kaita College of Education, Dutsin-Ma, Katsina State, Nigeria
Aliyu Dahiru Karofi Isa Kaita College of Education, Dutsin-Ma, Katsina State, Nigeria

DOI: https://doi.org/10.58709/niujed.v12i3.2715

Abstract

Integrated Science teacher education is central to developing scientifically literate learners capable of understanding and applying concepts across biology, chemistry, physics, and environmental science. However, teacher education programmes in many Nigerian contexts face persistent challenges, including limited instructional resources, inadequate pedagogical support, linguistic diversity, and uneven access to digital technologies. This paper presents the architecture and Python reference implementation of an adaptive, multilingual AI-powered pedagogical support framework for Integrated Science teacher education, designed for deployment in Nigerian colleges of education. The framework organizes classroom interaction as a three-stage input, processing and output pipeline. A Bloom's-Taxonomy-tagged instructional corpus feeds a Low-Rank Adaptation (LoRA)-fine-tuned LLaMA language model, teacher voice queries in Hausa are transcribed with a fine-tuned XLSR-Wav2Vec2 speech recognizer and translated for processing, and generated instructional content is delivered back to teachers as text and synthesized Hausa speech. Through voice-based and local-language interaction, the system promotes inclusive and context-responsive science teaching while embedding ethical safeguards. We situate this design against recent advances in African-language natural language processing and low-resource speech recognition, provide worked code for every pipeline stage, and outline an evaluation plan combining automatic speech-recognition and translation metrics with teacher-facing usability assessment across partner colleges of education. The framework is offered as a reusable, extensible architecture for AI-supported, multilingual teacher education.

Keywords: Integrated Science teacher education; artificial intelligence in education; low-resource natural language processing; Hausa speech recognition; Bloom's Taxonomy; multilingual pedagogy; Africa