Previous studies exploit the complementary effect between text and images in microblog sentiment analysis, but less focuse on some factors such as user personality and content characteristics. To solve this problem, this paper proposed an multimedia sentiment analysis on microblog based on multi-feature fusion method. Firstly it constructed a text sentiment classification model considering the user-based features and content-based features which had a good indication of emotions. Then it constructed an image model based on parameter transfer and fine-tuned. Finally, it designed fusion method of the early fusion and the late fusion. The experimental results show that user-based features and content-based features can enhance model ability of capturing emotion semantic, and the proposed model and fusion methods get higher accuracy on the real-world test dataset.

[2020-07-05] Printed Article /doc/2018.12.0929/v2