Bi-GRU Relation Extraction Model Based on Keywords Attention

Abstract Relational extraction plays an important role in the field of natural language processing to predict semantic relationships between entities a sentence. Currently, most models have typically utilized tools capture high-level features with attention mechanism mitigate adverse effects noise sentences for prediction results. However, task relational classification, these mechanisms do not take full advantage information some keywords which on expressions sentences. Therefore, we propose novel relation model based keywords, named Relation Extraction Based Keywords Attention (REKA). In particular, proposed makes use bi-directional GRU (Bi-GRU) reduce computation, obtain representation sentences, and extracts prior knowledge entity pair without any NLP tools. Besides calculation entity-pair similarity, REKA also utilizes linear-chain conditional random (CRF) combining features, similarity its hidden vectors, weight resulting from marginal distribution each word. Experiments demonstrate that approach can utilize incorporating expression semantics assistance achieve better performance than traditional methods.