Multimodal Sentiment Analysis of Social Media Based on Top-Layer Fusion

With the diversification of social media, users tend to express their feelings and opinions in the form of images and text. Compared with single-modal data, multimodal data can express users' feelings and sentiments more vividly and interestingly. Based on these, in this paper, we propose a top-layer fusion strategy and we construct a multimodal sentiment analysis model based on this strategy. The model can take into account the advantages of different fusion methods and can take full advantage of the superior performance of the single-modal sentiment analysis model. The experimental results on the MVSA-Single dataset show that the proposed model outperforms strong baseline models by large margins, demonstrating the effectiveness of the proposed model.