Entity Relation Extraction in Equipment Maintenance Support Based on Dual Attention Mechanism

Abstract

Aiming at the lack of labeled training data in the field of equipment maintenance support, a training instance set of entity relation extraction of equipment maintenance support is constructed by distant supervision method. Furthermore, aiming at the problem of noise data in the training instance set automatically constructed, an entity relation extraction model based on dual attention mechanism is proposed in this paper. The model adds character-level attention mechanism and instance-level attention mechanism after bidirectional gated recurrent unit (BI-GRU) network, which enables the model to obtain the bidirectional context semantic information of training instances through BI-GRU network, and focus on the key semantic information in each training instance through character-level attention mechanism. At the same time, the instance-level attention mechanism is introduced to calculate the correlation between the instance and the corresponding relation among multiple instances to dynamically reduce the weight of wrong label instances, so as to reduce the impact of noise data. The experiments on Chinese equipment maintenance support training instance set show that the model can effectively utilize the semantic information contained in instances, reduce the impact of the wrong label instances and improve the precision of relation extraction.