Sentiment Analysis For Specific Areas Of Social Networks

The social network has become an indispensable part of people's modern life and has gradually changed many traditional habits.People's way of reading news has moved from paper to online.News media can publish news comments through social networks.Meanwhile,the interactivity of social networks also provides a platform for all users in the network to express their opinions.With the development of the Internet,people can consume all the necessities of life through the Internet.To improve the user experience,the product review system is gradually improved.Both the news comment and the product comment contain the author's emotional tendency towards a certain thing.Therefore,this paper conducts emotional analysis and research on two major areas of social networks(news comment field and product comment field).For the study of news comments,this paper starts from the author and content of news comments in social network.Firstly,aiming at news media,the publishers of some official news comments,this paper proposes an event-level news media influence evaluation method.This method combines the network structure and user behavior of social networks,and the sudden social security incidents represented by the violent and terrorist incidents in xinjiang were studied.The study started with the authors of news reviews and quantified their impact.The experimental results show that news media in different countries have different influences on different events,and the same type of events have different influence scopes due to different places,which also reflects the differences in political positions of different countries.Secondly,starting from the content of news comments,this paper has finished the research on the position testing of news commentary.As a task of sentiment analysis,the main objective of position detection is to predict the position tendency of relevant comments on a particular topic.Most of the existing studies only consider the relationship between text and target words.This paper proposes a model based on knowledge learning and a model based on syntax learning.To fit the judgment habit of human beings,the former used the model to provide a series of common-sense knowledge to the machine to help predict the author's position on a certain topic.The experimental results show that using common sense knowledge can better learn the deeper implicit relationship.The latter deeply explores the semantic information in the text by learning the syntactic structure of the text.The experimental results show that syntactic learning plays a good role in the position detection task.Stance detection has good application value to the monitoring of public opinion.In the field of product comment,this paper chooses a new style of product comment,Q&A product comment,for emotional analysis.Q&A product comment is different from traditional product comments in that the answer has more specific relevance to the question.The sentiment analysis task of Q&A product comment can be regarded as a special aspect level emotion analysis task.And a BERT-based model is proposed in this paper.The model has been tested on both the Q&A dataset and the aspect-level sentiment analysis dataset,and the results show that this model has good performance and generalization ability.This study could help users learn more about the product,as well as help businesses upgrade their products.