Voice-First Mathematics Access for Visually Impaired Marathi-Medium Learners: A Mixed-Methods Evaluation of an Offline Tutor

Abstract

This study reports the design and evaluation of an offline, voice-first mathematics tutor for blind and low-vision (B/LV) Marathi-medium learners aged 11–14. Existing accommodations—sighted readers, generic screen readers, and digital audio textbooks—either compromise learner independence or assume English-medium interfaces and stable internet connectivity unavailable in rural Maharashtra. Grounded in Universal Design for Learning, the Expanded Core Curriculum, and sociocultural theories of mathematics learning, the system runs on a Raspberry Pi 4 and accepts spoken Marathi as its only input. A mixed-methods, three-arm matched-cohort design with 30 learners across four schools compared the proposed system against an audio textbook (control) and an earlier prototype (baseline). Over eight weeks, the intervention arm gained 40 percentage points on a 25-item oral mathematics assessment, versus +24 (baseline) and +12 (control); Cohen's d = 1.94 against control. Gains were largest in conceptual understanding, verbal explanation, and transfer. Learner acceptance averaged 4.6/5 and teacher acceptance 4.5/5. Findings indicate that voice-first, offline, mother-tongue tutoring can meaningfully advance inclusive mathematics in low-resource regional-language contexts.