Entity Relation Extraction Model Based Negative Feedback Attention

The relationship between entity pairs is the key to understanding an entire sentence or paragraph in a natural language, so entity relationship extraction is particularly important in natural language processing tasks. When we understand the information of a sentence, we often need to combine the information into the context. The entity relationship of a sentence is expressed as a relationship, but combined with the next sentence, the entity represents another relationship,and It called negative feedback. In this paper, to solve these problems, we propose NFBATT.a entity relationship extraction model based negative feedback attention.which is based on improved attention and pre-training of BERT.The main is highlighted the negative feedback in attention of Distant Supervision Relation extraction,and in order to get a better vector representation get a better vector representation of semantics, the sentence vector we represent is a combination of word Embedding, entity Embedding and positional embedding of solid and non-solid entity words by deep learning, which we call sentence multi-information representation. We use the negative feedback attention to reduce the influence of negative feedback on sentence relationship discrimination. At the same time, we have also done a lot of experiments to prove it. Through extensive experiments on Sogou Person Knowledge Map dataset.we demonstrate NFBATT's effectiveness.and the accuracy rate is 82.0%, which is better than the best method currently available.