Entity Relation Extraction Model Based on Semantic Information and Attention Mechanism

Entity relationship extraction is an important task in information extraction, aiming at extracting all entity relationship triads from unstructured text. However, existing entity-relationship extraction models have insufficient semantic information extraction features in extracting subjects and ignore the potential relationship between subjects and sentences in extracting relational objects, resulting in poor extraction triads. Therefore, this paper proposes a joint entity-relation extraction model based on semantic information and attention mechanism. Firstly, the sentence input is encoded by the RoBERTa-wwm model; secondly, the semantic information of the sentence is learned using BiLSTM as a feature enhancement layer; again, the subject extraction is implemented; then, the attention of the subject and the sentence is extracted and they are incorporated into the encoding vector; finally, the relationship and the object are extracted. The experimental results show that the model outperforms the baseline model on the DuIE 2.0 dataset, which verifies the validity of the model.