This output explicates the theoretical and methodological bases of the branching dialogue system that underpins the platform. It synthesizes CALL research (including evidence for medium-sized L2 gains in dialogue-based systems), addresses ASR limitations for beginners, and proposes a closed-domain recognition strategy that matches spoken input to predefined alternatives. The architecture—draw.io-based flow control with staged feedback (implicit→explicit, forced progression after repeated errors), acceptance of alternative correct responses, and standardized .docx video scripts—is implemented through an iterative prototype→pilot→revision cycle to ensure pedagogical validity and technical reliability.