Research And Application On Automatic Summary Algorithm Based On Multiple Models

With the popularity of the Internet and the increase in network users,the amount of information on the network has grown exponentially.In this era of big data,while the Internet brings us convenience,its data overload problem also brings challenges to intelligence analysis and other tasks.In order to improve the information conversion rate of intelligence workers and discover the "long tail" in the information,this paper proposes a single-text summary algorithm for military news and a multi-text summary system based on semantic analysis.The single-text summary algorithm summarizes the text content and speeds up the reading of military articles;the multi-text summary system summarizes the article,optimized from the space and reducing the working of intelligence workers.The text summary system alleviates the problem of information overload in the network.Therefore,the automatic summary algorithm is a research direction of great significance.This paper proposes an automatic text summarization model construction method based on BERT Embedding.This method based on the BERT pre-training language model.The model is used to enhance the semantic representation of the word vector.The generated word vectors are input into the seq2seq model for training and forms an automatic text summarization model to rapid generation of the text summarization.This model can effectively improve the accuracy and readability of generating summaries on Gigaword,CSL and other datasets,and can be used for automatic text summary generation tasks.Then,based on the automatic abstract model proposed in this paper,combined with semantic analysis related technologies,a topic-oriented multi-text abstract system is proposed.The LDA topic model and AP clustering algorithm are used to extract the topics and events from the text.The result of the topics and events is combined with the single text summary algorithm to calculate the event-oriented multi-text summary.This paper obtains data from the Internet for experiments and analyzes the results to verify the feasibility and effectiveness of the multi-text summary system.