| Along with the rapid development of chemical industry,chemical accidents also happen frequently.As there are many inflammable,explosive,toxic and corrosive substances in chemical products,once improper management or mistakes in production,accidents such as fire,explosion,leakage or poisoning may occur.In order to avoid and reduce the occurrence of accidents and the losses caused by accidents to the greatest extent,it is necessary to know which behaviors,goods or events cause chemical accidents.Therefore,it is of great significance to classify the text of chemical accidents to analyze the causes of accidents,take preventive measures,and supervise,warn and deal with chemical accidents.In addition,with the continuous increase of the speed of network information transmission,online public opinions,especially microblog online public opinions,have extremely fast transmission speed and strong influence.Once a big chemical accident happens,it is bound to arouse wide attention from all walks of life.Therefore,how to grasp the public opinion in time and guide it correctly has become an important work.Based on the above questions,this paper studies the text classification of accident news in the chemical industry and the analysis of public opinion on weibo and the Internet.The main research work is as follows:1.Crawling and preprocessing of texts related to chemical accidents.Web crawler technology is used to crawl relevant chemical accident news text and accident comment information on major websites and microblogs.Subject model,similarity algorithm and search keywords are used to filter and remove noise data with low or no correlation with chemical accident.After manual labeling,word segmentation and statistical methods are used to preprocess the data.2.Feature extraction and text classification of chemical accident texts.In this paper,a text classification method of chemical accidents based on BLSTM-Attention neural network is proposed.This method utilizes the Bi-directional Long Short-Term Memory(BLSTM)method to estimate the features of chemical accident news text,which can fully learn the sequence information of chemical accident text context and effectively express the semantic information of chemical accident text.The introduced Attention mechanism assigns corresponding weights to different words and sentences according to the different contributions of words and sentences to the article,and retains the feature information that is more valuable and meaningful to the text,use the softmax function to complete the classification of the accident and output the result.The introduced Attention mechanism assigns corresponding weights to different words and sentences according to their different contributions to the article,retains feature information that is more valuable and meaningful to the text,and USES softmax function to complete the category determination of accidents and output the results.The experimental results show that the text classification method of chemical accident based on BLSTM-Attention neural network proposed in this paper is more advantageous in the text classification task of chemical accident news,and its classification accuracy can reach 92.12%.3.Analysis of public opinion on chemical accident network.According to the classification results of chemical accidents,the author analyzes the public opinions on weibo and Internet for the especially serious chemical accidents.The potential semantic index LSI text similarity index is used to study the weibo data of chemical accidents and obtain the trend of accidents.Using LDA theme model to mine the potential emotional theme tendency of microblog comment data;Based on THE LSTM neural network neural network method,the emotional preference classification task of public opinions was completed,and combined with the classification results of chemical accidents,accident hazard degree,topic discussion and other data,the comprehensive analysis of the accident's microblog network public opinions was carried out. |
PDF Full Text Request