Multi-feature Fusion for Relation Extraction using Entity Types and Word Dependencies

Most existing methods do not make full use of different types information sources to extract effective features for relation extraction. This paper proposes a multi-feature fusion model based on raw input sentences and external knowledge sources, which deeply integrates diverse lexical, semantic, syntactic into deep neural network models. Specifically, our extracts lexical granularity from the original text representation, entity type annotation corpus, dependency trees. Meanwhile, dimension-based attention mechanism is proposed enrich diversity enhance their discriminability. Different enable comprehensively utilize various information, so this fuses these train classifier The experimental results show that outperforms state-of-the-art baselines TACRED Revisited, Re-TACRED, SemEval datasets, with macro-average F1 scores 81.2%, 90.2%, 89.4%, respectively, improving performance by 1.4%, 4.4%, 2% average, indicates effectiveness modeling.