Research And Application Of Sentiment Analysis Based On Network Short Text

Individual users represented by We Media have overtaken traditional mainstream social media as the main body of information distribution in recent years,thanks to the rapid expansion of Internet sectors such as online social networks and e-commerce platforms.The network platform has a huge number of remark texts with subjective emotions.Analyzing and mining opinions and emotional information from large texts,as well as comprehending people’s emotional reactions to public events,can help governments and businesses design suitable response measures in a timely manner.Furthermore,it is critical for ensuring cyberspace security.What is the best way to conduct sentiment analysis on natural language processing and other technologies have steadily drawn the attention of more and more academics to network short texts.Aspect-based sentiment analysis,as opposed to gengral sentiment categorization,focuses on more fine-grained sentiment information.It takes advantage of the sentiment link between aspect words and context to anticipate the sentiment polarity of a particular aspect in context.Aspect-based sentiment analysis has recently seen a surge in the application of graph convolutional neural networks based on syntactical dependency trees.The majority of previous research,on the other hand,focuses on the syntactic relationships between contextual words and the aspect aspect,ignoring the use of affective commonsense knowledge.Furthermore,network short texts,such as product reviews,are brief.The graph convolutional neural network sentiment analysis model can’t improve since the content and structure of these texts are excessively subjective and arbitrary.Therefore,a dual graph convolutional sentiment analysis model(Dual Graph Convolutional Network for Aspect-based Sentiment Analysis,DGCN ABSA)is presented to solve these issues in aspect-based sentiment analysis.To extract different text features,it employs two graph convolutional networks.To begin,a Sak_GCN module based on syntactic and emotional information is proposed,which employs SenticNet’s affective knowledge to improve the syntactic graph of sentences.It also extracts affective information from context and aspect words,as well as connections between them.Second,to extract semantic correlations across contexts,a Sat_GCN module based on self-attention is proposed.Then,for sentiment classification,swap and combine features from these modules.To evaluate the model’s performance on aspect-level sentiment analysis tasks,a series of experiments are undertaken on five public datasets.An aspect-based sentiment analysis system for short texts is created and built based on the preceding sentiment analysis research.The system has two key functions:data analysis and sentiment analysis based on aspects.The data analysis function extracts the inherent relationship between data and visualizes it.The sentiment polarity of a user’s provided aspect in the defined text is predicted by the aspect-based sentiment analysis tool.