Context-Based Visual Sentiment Analysis for Social Media Data

Abstract

Social networking sites have recently grown in importance and popularity, so the field of textual sentiment analysis has emerged and attracted a great deal of research interest, additionally, sentiment analysis in images is still in its infancy and little research has been conducted in this area; listing text or visual content alone is insufficient to convey And the opposite of the feelings of the published content; therefore, it was proposed to analyze the visual feelings based on the content of the image. In this paper, a system was proposed to determine the polarity of posts and tweets on social networking sites using textual analysis and visual analysis. A system or model was proposed that integrates these different properties using a proposed neural network (DVSF) that integrates the text model and the visual model to provide a final decision to indicate the polarity of these posts. Twitter (MVSA) and Flickr(EmotionROI) datasets were utilized, and the results were encouraging overall.