Research And Implementation Of News Event Analysis System Based On Deep Learning

With the development of the Internet,the amount of information in the Internet is rapidly increasing at an exponential level,and network news has become an indispensable source for people to obtain information.The existence of a large amount of redundant information in the network makes it difficult for users to have an ins and outs of a specific event on the news media.To deal with these challenges posed by the information explosion,we need some automated analysis tools to help people find truly valuable information.Event detection can gather highly relevant news information together,which can enable people to obtain information more efficiently,catch up social trends,and not be overwhelmed by a large amount of redundant information.Event prediction captures information in pre-order sub-events to automatically generate next predicted events,which is of great significance to government,companies and individuals.This article introduces deep learning technology into news events analysis.The main research work is as follows:(1)This paper proposes an event detection algorithm that incorporates structured information from news.The main contribution of this algorithm lies in incorporating news timestamp information into news representation learning,and designing a two-level cluster event detection model based on key named entities.This method not only emphasizes the role of structured information in news,but also overcomes the disadvantages of traditional clustering methods,such as K-means,that need to set the number of clusters.And the model is evaluated on the crawled news corpus,which is significantly better than the state-of-the-art model.(2)This paper proposes an event prediction algorithm,named DRAM,which utilizes attention mechanism and reinforcement learning.The algorithm is based on sequence to sequence learning framework,and a novel layered attention mechanism is designed according to the news sequence structure,and deep reinforcement learning is introduced into the event prediction task.The advantage of this method is that the model can automatically evaluate the quality of the predicted events and motivate itself to generate high-quality predicted events.Experiments on real data sets show that our model is significantly better than other state-of-the-art model.(3)This article also builds an integrated news analysis system.The above two points both improve the effect of news analysis from the algorithm level.In order to facilitate users to quickly and accurately analyze news data sets,this article builds an integrated news analysis system based on the Spring MVC framework,which can receive news data set uploaded by users and automated anlysis news set.The main functions include user management,data exchange,event detection and event prediction.