Geological Relation Extraction Method Based on Attention Mechanism and PCNN
Relation extraction in the geological domain is of great significance to the intelligent analysis of geological data and the construction of knowledge graph of Geology.To solve the problems of missing data sets and the difficulty in feature extraction for relation extraction tasks of the geological domain,this paper builds a large-scale relation extraction data set in geological domain based on the idea of distant supervision,and proposes a distant supervision relation extraction model which combines attention mechanism and piecewise convolutional neural network(PCNN).The model in this paper uses the piecewise convolutional neural network to automatically extract the semantic features of the training instances.At the same time,the segmented attention mecha-nism is added to the piecewise convolutional neural network to highlight important segments in the instance,which strengthens the model's ability to extract important features in training instances.In addition,a sentence-level attention mechanism is introduced in-to this model to reduce the impact of incorrectly labeled data due to the distant supervision.The experimental results on the relation extraction data set in the geological field based on distant supervision show that the precision,recall,and F1 value of this model are all higher than other baseline models,which illustrates that the method has better relation extraction capabilities.
国家科技期刊平台
NSTL
万方数据
