Few-Shot Relation Extraction via the Entity Feature Enhancement and Attention-Based Prototypical Network

Figure 1

The overall architecture of the proposed model. It consists of a RoBERTa-based encoder, a BiLSTM-based directional feature extraction module, an entity feature enhancement module, and a relation prediction model via attention-based prototypical networks.