bert-topic

Proposes SSDAU, a structured semantic data augmentation method for joint entity and relation extraction that preserves semantic structure by segmenting text based on entity labels and using BERTTopic for topic consistency, significantly outperforming existing augmentation methods.