A Chinese L2 Learners' Dynamic Vocabulary Growth Network Model Based on Graph Deep Learning

This paper regards vocabulary networks mastered by Chinese second language(L2) learners at different levels as sub graphs of a Chinese Word Co-occurrence Network, embeds these subgraphs with the help of graph deep learning techniques such as TSPMiner model and Order Embedding algorithm, and builds a dynamic vocabulary growth network model for the learners. This model can predict nodes and links between nodes, simulate the growth process of a learner vocabulary, so as to offer guidance to learners. With this model, a smooth, efficient, and dynamic adaptive vocabulary learning process becomes possible on learning platforms. Through a questionnaire and data analysis on it, the model is verified in that participating Chinese teachers have great consistency with model recommended word learning sequences.