Research On Text Extraction Method Based On Key Sentence And Keyword Association

With the popularization and rapid development of the Internet,information data is also growing rapidly.The Internet is full of a variety of different formats and types of data,such as news,blogs,forums,posts,etc.On the one hand,a large number of data enriches our lives,on the other hand,it also poses a challenge to modern technology.For example,how to extract effective information needed by users from massive data.To solve this problem,various technical methods follow,keyword extraction technology and text summary sentence extraction technology are two of the solutions.This paper proposes two hypotheses from the perspective of the connection between key sentences and keywords.One is that the words in the key sentences have a high probability of being the keywords of the text,and the other is that the sentences appearing in the keyword set have a high probability of being the key sentences of the text.The relationship between the two is closely related,and it can also provide new ideas for keyword extraction technology and text summarization sentence extraction technology.Based on this idea,the work of this paper is summarized as follows:First,for keyword extraction technology,this paper proposes to use two supervised summary sentence extraction methods based on Bert model and graph model to extract key sentences in the text data,and then extract the key information of the text,and then perform relevant text processing operations such as reordering and deleting redundant tags on the extracted key sentences.Finally,the new key sentence text data set is put into the supervised DeepCT keyword extraction method to extract the text keywords and weights.The experimental results show that the optimized key sentence data set can improve the effect of text keyword extraction method.Second,for text summarization sentence extraction technology,the purpose of this technology is to extract key sentences from the text.Firstly,the keywords and their weights of the text are obtained by the keyword extraction method.Then,the obtained keywords and their word weights are integrated into the construction of word nodes and edge weights of the graph model,and irrelevant sentence nodes are removed according to the keywords,thus optimizing the initialization structure of graph nodes.Finally,the key sentences of the text are obtained through the iterative update of the graph attention network and the sentence selector.The experimental results show that incorporating keyword features can improve the effect of text summary sentence extraction method.The experimental results show that there is a certain mutual promotion between key sentences and keywords.Adding key sentences into the keyword extraction method and incorporating keywords into the text summary sentence extraction method can promote the extraction effect of each other.