KEIC: A tag recommendation framework with knowledge enhancement and interclass correlation

Tag recommendation is critical in organizing and managing resources on social media platforms. The incessant deluge of new content and terms and the challenge of creating new tags significantly complicates this task. To address these challenges, we present KEIC, a tag recommendation framework that unifies Knowledge Enhancement and Interclass Correlation. KEIC enriches the deep semantic understanding of the text by incorporating commonsense knowledge and identifies interclass correlations to migrate the long-tail effect of tags. Experiments conducted on three large-scale datasets have demonstrated that KEIC can seamlessly integrate with existing classification-based tag recommendation models, delivering outstanding performance without the need for excessive parameter augmentation, while effectively mitigating the long-tail effects in the data. The ablation experiments further confirm the effectiveness of different components of our proposed framework.