KeyHierQC: Leveraging Keywords and Hierarchical Data for Enhanced Classification of Question Knowledge Points

Efficiently managing online question resources is crucial for intelligent education systems. Correctly sorting questions into hierarchical knowledge points enhances educational content’s accessibility and effectiveness. Many existing classification methods fail to address the inherent label hierarchy and the scarcity of semantics in question texts, leading to inadequate results. Therefore, we propose the KeyHierQC model, utilizing keywords and hierarchical data to improve question knowledge points classification. KeyHierQC incorporates the Hierarchical Keywords Attention Mechanism (HKAM) and Semantic-aware Contrastive Learning (SCL) to better process short texts and link detailed semantics to appropriate labels. Finally, our extensive experiments on two problem classification datasets show the efficacy of our proposed method and indicate the necessity of incorporating keywords in QC tasks.