In our country, many elementary school children go to tuition classes after school. Mathematics is one of the most frequently attended tuition subjects because children do not understand all of the concepts covered in textbooks during class time. Compared with the elementary school mathematics textbooks in other countries, our mathematics textbooks cover more units within the same teaching time, based on the assumption that children like novel things. However, there seems not to be enough time for children to familiarize themselves with a new unit, and the degree of cohesion between units needs to be adjusted. In this study, we collected 6,674 math word problems from three versions of textbooks, covering twelve semesters, and tried to cluster these problems using Latent Semantic Analysis in Python. Unlike text classification, where a text is classified into a pre-defined set of classes, the text mining technique used in this study identifies co-occurring keywords to discover hidden topics in the collections of textual information. It is an unsupervised text analytics algorithm that is used for clustering documents. A model with a high topic coherence score will be considered a good topic model. At the end of this study, a new order of teaching units is proposed based on the hidden topics suggested in the best topic model, in the hopes of providing a scientific basis for scaffolding by breaking up teaching units into clusters so that schoolchildren can grasp new materials and retain what they've learned in the same semester.
Ting-Hao Yang, National Tsing Hua University, Taiwan
Ching-Ching Lu, National Tsing Hua University, Taiwan
Wen-Lian Hsu, Asia University, Taiwan