Gamification of Japanese temple geometry in chatbot for advanced personalized learning

Abstract

In this Final Year Project, we present a personalized geometry learning platform for students to develop spacial ability through solving geometric construction puzzles. Geometric construction problems refer to a set of construction tasks where users are prompt to use tools such as straightedges and compasses to construct geometric shapes. The system we propose consists of several modules. The first module is a conversational agent in form of an embedded Messenger chatbot. The chatbot is able to interact with the user with human traits, and attract students to work on the problems. The second module is a geometric construction problem generation module implemented with two groups of Generative Adversarial Networks (GANs). The module is able to generate non-existent geometric construction problems of various difficulty levels. The third module is a geometry solver implemented with Mask R-CNN. This module is able to solve geometric construction problems through images without textual inputs, and is used to determine whether a GANgenerated problem is a valid solvable problem. Apart from these components, we also integrate two advanced algorithms: the Iterative Deepening Search Algorithm (IDS) and the Divide-and-Conquer Algorithm. With IDS, the process of searching for steps can be conducted more effectively. With divide-and-conquer, the geometry solver based on Mask R-CNN can solve rather complex problems like Japanese Sangaku instead of just trivial ones. Combined together, all components form an integrated dialogue system where students can practice at their own pace, and learn geometry through constructions. As such, the system is believed to be able to boost advanced personalized learning for geometry education.