The Research And Implementation Of Text Summarization Based On Seq2seq Framework

With the explosive growth of information in the new era,people could get and share information from various sources.The network contains millions of data and documents,it is growing at an exponential rate.Therefore,we are facing the inevitable and challenging problem of information overload.Automatic summarization can alleviate this problem.These concerns have caused people's interest in the development of text summarization systems.Text abstracts are designed to take a single document or group of documents as input and produce a concise summary that include the most important information.Text summary of long text data in the world of big data has great significance for people to be able to quickly and accurately get important data from massive data.Most of the previous text abstracts focused on short text data,but the lack of long text s is difficult to meet the requirements of the current big data era.On the basis of the analysis and summary of the text vector representation and the principle of machine learning model GRU,this paper studies the problem of using machine learning model to solve the problem of text summarization.The main research work in this paper is as follows:(1)For long text sequences,the semantic expression of the whole text sequence is expressed only by the expression of an intermediate semantic vector C,which will have obvious loss of feature,and the semantic information and details of the text sequence itself may have disappeared.This paper designs a multi-layer encoder decoder model using word sentence paragraphs,which provides a method for long text summarization.(2)The importance and relevance between text are calculated based on the traditional graph sorting method,this method is used to calculate the weight distribution of each semantic vector after the coding,and then the improvement is made on the basis of this method.The output of the text sequence will also affect the weight of subsequent text sequences.(3)Aiming at the forward dependence problem of GRU model,this paper designs a combined positive reverse sequence GRU model,which combines the vector of the positive reverse sequence as the feature vector,and uses the Bi-GRU model as the contrast model to explore the influence of context environment of text context to text summary.Finally,by comparing the 4 text summarization models,we analyze the Bi-GRU multilayer model based on the improved graph attention mechanism.Imporve the public opinion analysis system,automatically generate text summaries from the network data captured by the public opinion analysis system,and display them in the public opinion analysis system.