Classification technique of appeal text based on ERNIE-AA model
12345 citizen service hotline brings together a large number of citizen appeals. Classification of these appeal texts can effec-tively improve the efficiency of citizen service hotline. In order to classify the appeal text quickly and accurately, this paper proposes an ERNIE-AA model based on artificial attention mechanism. The model introduces the heuristic rules of artificial classification into the input representation of ERNIE model by artificial attention mechanism, and increases the attention weight of keywords that affect classi-fication by pretraining model. The experimental results show that the ERNIE-AA model has good performance on the appeal text data set, and the classification accuracy can reach 86. 71 ％.
万方数据
维普
