Research On Topic Classification And Information Extraction For Public Opinion Monitoring In Animal Health Field

With the development of economic globalization,the circulation of animals and their products become more and more frequent,and the chance of disease transmission increases.Combined with environmental pollution and ecological balance destruction,the animal pathogen has been mutated,which brings difficulties to the prevention and control of diseases.Nowadays,with the rapid development of Internet,the Internet has become the main way for people to communicate and get information.More and more people are accustomed to express their opinions and opinions on the Internet.Once the animal health events occurs,they will attract the attention of a large number of people.China is becoming a rare magnetic field of animal health public opinions in the world.If we don't effective guidance and manage the public opinions,it may cause the loss of the enterprises or harm the social stability order.Therefore,how to monitor and analyze the animal health-related public opinions has become the key point of the research.In this paper,we use news text as the research object to do the research on the animal health-related public opinions.Based on the full communication with China Animal Health and Epidemiology Center,we comprehensively investigated the related documents,methods and technologies at home and abroad,and proposed using topic classification and information extraction methods to research animal health-related public opinions information.Firstly,we analyze the source,category and description characteristics of animal health-related news,and add the text features and animal disease attributes to the weighted Naive Bayes classification process to build a classification model suitable for animal health-related public opinions information.Secondly,based on summarizing the description feature,direction and position of attribute,we take animal health safety public opinions information as an example,use regular expressions to build extraction rules to achieve the extraction of time,place,epidemic disease,number of animals,and measures.Moreover,we propose an extraction method based on the comparison of node relationships to improve the number attribute items of animals with poor extraction efficiency.Finally,based on the research results of classification and information extraction,combined with the actual needs,we realized the prototype design of animal health-related public opinions monitoring system on B/S architecture.