Design And Implementation Of News Title Generation System Of Party Building Based On Automatic Text Summarization

With the rapid development of mobile Internet,more and more party building information is more inclined to be published on official microblog,friend circle,Party building app and other Internet network platforms.People are surrounded by massive text information in daily life and work life.How to get the content that people need quickly and accurately from the massive text information has become a very urgent demand in today's society.Automatic text summarization is a sub task in the field of natural language processing,which can transform the original long text document into a short summary description.At present,there are two main methods of automatic text summarization,one is extraction text summarization.This method is mainly based on linguistics,focusing on the lexical,grammatical and semantic information of the text.By reordering the sentences of the original text according to the importance algorithm,the most important sentences are extracted and combined into a paragraph of summary.Although the method of extracting sentences as abstracts is easy to implement and widely used in technology,the abstracts generated are not concise,coherent and comprehensive.The other is the generative text summarization,which uses the most popular deep learning method at present,and uses the neural network to learn the text features and generate the summarization automatically.Combining the two methods,this paper proposes a TR-MA-PGN algorithm model for Party building news data.The work mainly focuses on the technology of text summarization in the field of Party building news,so it collects and arranges the special corpus of Party building news.The overall architecture of the system mainly includes data processing module,algorithm model module and application service module.The TR-MA-PGN algorithm model is the core pillar of the system.Based on the two-way cyclic network,the system algorithm model adds the multi-terminal self attention mechanism to extract the text feature information from the multi-level.The introduction of pointer network and coverage mechanism effectively improves the quality and readability of news headlines.Finally,the trained algorithm model provides the API interface,which is called by the portal website and convenient for people to use.The TR-MA-PGN algorithm proposed in this paper is more readable and comprehensive than the summary generated by other benchmark models,and the feasibility of the party building news title system is verified through the functional test of the system.