DUAL-CHANNEL TEXT SUMMARIZATION GENERATION METHOD BASED ON GRAPH ATTENTION
The traditional extraction summarization method fails to fully consider the different meanings between character features and word features in the Chinese environment,and it is easy to misunderstand the information of the original text.In this paper,a dual-channel Chinese text summarization generation method based on graph attention is proposed.The two-channel encoder structure was used to extract the character-level and word-level characteristics of text respectively.The method built a three-way structure for each text and extracted the characteristics of triplet sets corresponding to the text through the graph attention.The feature fusion was entered into the decoder with the copy mechanism for summarization generation.The method could effectively improve the information extraction ability of the original text.The results of the comparison experiments show that the method performs better on both datasets.
Text summarizationAttention mechanismAbstractive summarization
国家科技期刊平台
NSTL
万方数据
维普
