SENTIMENT ANALYSIS IN SOCIAL MEDIA USING MULTI-MODAL DATA FUSION TECHNIQUES

Abstract

In recent years, with the popularity of social media, users are increasingly keen to express their feelings and opinions in the form of pictures and text, which makes multimodal data with text and pictures the con tent type with the most growth. The bulk of information written on the social media by users is clearly sentimental in nature and multimodal sentiment analysis has emerged as a vital area of study. Receiving the text and image feature separately then integrating them to classify the sentiment has been the main focus of previous studies on multimodal sentiment analysis. These researches tend to overlook the interrelationship between text and images. Thus, the new model of multimodal sentiment analysis is suggested in this paper. The model initially removes noise interference in text related information and cleanses more significant image features. After that, under the feature-fusion step which is founded on the attention mechanism, the text and images teach the inner features of each other with the help of symmetry. Then the fusion features are implemented on sentiment classification problems. The proposed model proves to be effective as evident in the experimental findings of the two popular multimodal sentiment datasets.