Research On Text Abstract Generation Method Based On Deep Neural Network

With the explosive growth of data information on the Internet,reading text as the main way to obtain information takes a lot of time to eliminate useless data.By extracting the text information summary,modern people can greatly improve the efficiency of understanding the original text,and effectively reduce the time and effort to obtain information.The main shortcomings of traditional automatic text summarization methods are extracting the combination of original sentences as a summary,semantic incoherence,and redundant content generated by feature extraction.Deep learning-based methods can effectively solve the disadvantages of traditional methods through the ability of neural networks to fit complex features.First,the traditional text abstract extraction methods and neural network-based text abstract generation methods are studied.By means of pooled sentence semantic representation,the fusion of language features improves the ability of sentence semantic representation.A bidirectional recurrent neural network text abstract generation model based on attention mechanism was realized,and the evaluation of automatic text abstract generation was improved.Secondly,the method of graph-based text summarization is studied.Through the construction of hierarchical bidirectional recurrent neural network,the semantic representation of text is realized.Drawing on the idea of the extractive abstract generation method,based on the traditional attention mechanism combined with the characteristics of automatic text summary tasks,we focus on the specific content of the text by establishing a graph-based attention mechanism.A graph-based decentralization mechanism was proposed.By traversing the global information of the chapter and optimizing the abstract generation,the evaluation of text automatic abstract generation was improved.Finally,a text summary generation system based on deep learning is designed and implemented.Through the modular design,a friendly human-machine interactive graphical interface is realized.Compared with the traditional method based on text information extraction methods,the results show that deep learning based The automatic text summarization method can generate higher-scoring text summaries.