Design And Implementation Of Cross-Platform Network Public Opinion Event Evolution Analysis System

In recent years,the number of netizens has been increasing day by day,and people can share their daily lives,express their opinions and communicate with others on social media platforms anytime and anywhere.More and more incidents,which are exposed on the internet and quickly generate heated debates among users all over the web,are gaining high attention on multiple platforms.The freedom of expression on the internet has contributed to the development of public opinion incidents,which seriously endanger the good online environment.For public opinion events with a wide range of influence and a deep degree of harm,it is necessary to find them early,pay attention to them,and guide and control them when necessary.Most of the existing analyses for public opinion events have problems such as single data source,ignoring semantic information based on clustering and lagging,which cannot achieve a effect of comprehensice,real-time and efficient analysis of events,and this is not good for a strong support for public opinion governance.In order to solve the problem of a single data source for analysis of public opinion events,this paper roughly divides the existing web platforms into four categories,namely university forums,public web forums,news websites and social media,and selects representative platforms from each category,such as Beiyang People’s Forum,Blog China,Renmin University of China News Network,CCTV,Weibo and Zhihu,etc.,and designs and implements personalized crawlers for different analysis sources to complete the work of classifying and collecting data from different analysis sources.In order to achieve cross-platform homologous event detection and recognition,this paper proposes an event recognition and analysis model based on graph convolution and semantic elements.This model completes pre-processing of data through text processing techniques,fuses word vectors and location vectors,vectorizes the data,and use a bi-directional long-and short-term memory network model to capture long-range contextual dependencies,and finally uses graph convolutional neural networks for feature enhancement and feature-based graph node classification to detect and recognise events in real-time data streams.This paper designs an event evolution analysis model based on semantic element extraction,distinguishes event types,divides time windows,extracts semantic elements from data within each time window using TextRank,and builds an exponential smoothing model to predict the heat of events using text spread and user participation.Based on the software engineering development process,this paper analyzes the requirements of the cross-platform network opinion event evolution analysis system and completes the system architecture design and system outline design based on the results of the requirements analysis.The system is divided into data collection and storage module,event detection and identification module,event evolution analysis module,task management module and visualization module,and the above functional modules are designed and implemented in detail;finally,system testing is carried out to verify that the system meets user requirements.