A multi-module approach to evaluate online teaching quality in international Chinese education

Abstract

The Quality of Online Teaching in International Chinese Education (OTICE) introduces a cutting-edge approach to distance learning, making educational content accessible without limitations related to age, gender, ethnicity, or location. This research aims to establish a robust evaluation framework with high predictive accuracy for assessing OTICE by leveraging ensemble and deep learning techniques. The study explores key questions surrounding sentiment analysis within educational data. Initially, we design an index system and determine evaluation based an online questionnaires framework for OTICE, while simultaneously compiling online data for corpus development. Subsequently, we create the Multi-Module Architecture Driven Model (MMADM), which integrates a 3D-CNN module, a gated mechanism, and a selection module. Across all evaluated setups, combining a gated mechanism with Bag of Words (BoW) and a Word2Vector (W2V) word-embedding model based on the skip-gram approach delivers the highest predictive performance. Empirical findings confirm that deep learning models outperform ensemble learning techniques in the context of educational data mining. Moreover, comparative model analysis reveals that the 3D-CNN module paired with the gated mechanism produces optimal results, achieving precision (P) and F1 scores of 97.91% and 97.90%, respectively. Compared to other models, the overall performance improves by 3% to 5%. These findings underscore the superiority of the proposed model in addressing the OTICE standard task objectives presented in this paper.