Research And Implementation Of Key Technologies In Sentiment Analysis Of Weibo Topics

At present,the number of Weibo users has reached 337.41 millions,and the number of posts on Weibo can reach hundreds of millions each day.Weibo has become the most influential one of social platforms nowadays.Every topic in Weibo triggers a lot of spontaneous discussions among users.Weibo users express their personal sentimental opinions through various interactions.Therefore,the research on the key technologies of sentiment analysis of Weibo topics can help us to be informed of the popular news and focus on the public sentiment.This thesis studies relevant key technologies of topic sentiment analysis,and designs a prototype system for sentiment analysis of Weibo topics.The main research contents of the thesis are as follows:(1)Aiming at the problem that the Single-Pass algorithm is sensitive to the input order of the text,a clustering algorithm called DPSingle-Passis is proposed to improve the selection method of the clustering center which is based on density peak in this thesis.This thesis defines the weight of the clustering center.In this thesis,the point with the largest weight is selected as the initial clustering center of the cluster.The experimental results show that the improved clustering algorithm can effectively improve the accuracy and recall rate of topic clustering.The average accuracy rate is 8.3% higher than the traditional Single-Pass algorithm,and the average recall rate is increased by 2%.(2)In consideration of the unapparent inclination of topic sentiment,a topic impact model TIM is proposed to quantify the sentiment value of the Weibo topics in this thesis.The model selects four attributes as the parameters of the model,which are audience coverage,user activity level,diffusion impact and the similarity extent between the content of the text and the topic.It can reduce the influence of anonymous instigators online on emotional evolution effectively.In this thesis,the SVM classifier is used to judge the emotional polarity of the text,and the emotion value is calculated with the TIM model.This method realizes the transformation of topical sentiment analysis from qualitative analysis to quantitative analysis.The experiment results show that the analysis of sentiment would be better and the sentiment topics would be identified more effectively with this model.(3)A microblog topic sentiment analysis prototype system is designed and implemented.This system analyzes the functional requirements in detail,and realizes the functions of microblog data collection,preprocessing,topic clustering and sentiment analysis.The prototype system verifies the effectiveness of the improved algorithm proposed in this thesis.