Public Opinion Robot System Based On Micro-blog

The research on Weibo,a social platform,has been a popular research direction in Chinese sentiment analysis and image-related text generation.It is a difficult point in Chinese word segmentation,Chinese text sentiment analysis,machine translation,and Image Caption.This article combines two kinds of technologies,the sentiment analysis of Weibo text and text generation related to images.At the same time,a public opinion-based robot system based on Weibo is implemented through a network communication protocol.The system has realized displaying the emotional trends of microblogs on specific topics,generating topic-based image descriptions,and simulating microblog functions such as login,sending,and viewing.The key issue of the public opinion robot system is to display the trend of the Weibo content of related topics,and at the same time,to control the hot topics by publishing texts.In order to realize the related functions of the public opinion robot system,this paper starts from the sentiment analysis of Weibo content and the text generation which related to the specific topic and divides the system into two major modules:sentiment analysis module and text generation module.The former compares the content of microblogs(including texts,pictures,and expressions)in related topics with different contributions to the text's overall emotion,sets different weights,and thus obtains the emotional tendencies of each microblog,and organizes and displays the statistics to obtain the trend of public opinion of this topic;the latter uses the convolutional neural network and memory network structure to generate texts related to topics for the images,and realizes automatic operations of generation and posting through network communication protocols and account control.The main contents and innovations are as follows:1.According to the content contained in each microblog,the sentiment tendencies are calculated separately,and the sentiment tendencies of a single microblog are calculated according to the different weights of different contents for the sentiment analysis.2.According to a specific topic,first retrieve the topics related to the picture,and then based on the picture,through the convolutional neural network and memory network model,predict text for each image related to a specific topic.3.Through the operation of simulating logging in microblog and network request operation,the system achieves retrieving the images,generating texts and posting microblog automatically.At the same time,the effective duration of different cookies is set for the account,so as to achieve simultaneous operation of multiple accounts.