To improve the accuracy of extraction task of aspect sentiment triplet,an extraction model based on joint dependency syntactic relations and location deviation information was proposed.The syntactic relation was added to the model context coding,and the structure and node attribute information were captured through the graph convolutional network to enhance the interaction ability between triplet elements.The relative position offset information was added to the multi-task learning part,and the relationship between aspect-opinion word pairs was fully mined,so as to improve the accuracy of triplet element extrac-tion.Experimental results on four standard English data sets show that the proposed method is effective and superior to other baseline models.
维普
万方数据