Sequence Generation Model Integrating Domain Ontology for Mathematical question tagging

Abstract

In online learning systems, tagging knowledge points for questions is a fundamental task. Automatic tagging technology uses intelligent algorithms to automatically tag knowledge points for questions to reduce manpower and time costs. However, the current knowledge point tagging technology cannot satisfy the situation that mathematics questions often involve a variable number of knowledge points, lacks the consideration of the characteristics of the mathematics field, and ignores the internal connection between knowledge points. To address the above issues, we propose a Sequence Generation Model Integrating Domain Ontology for Mathematical question tagging (SOMPT). SOMPT performs data augmentation for text and then obtains intermediate text based on domain ontology replacement to facilitate deep learning model to understand mathematical question text. SOMPT is able to obtain dynamic word vector embedding to optimize the textual representation for math questions. What’s more, our model can capture the relationship between tags to generate knowledge points more accurately in the way of sequence generation. The comparative experimental results show that our proposed model has an excellent tagging ability for mathematical questions. Moreover, the sequence generation module in SOMPT can be applied on other multi-label classification tasks and be on par with the state-of-the-art performance models.