关键词: 舆情分析, 情感分析, 特征提取, 图像描述, 多模态, 神经网络

Abstract: Public sentiment analysis is widely used in information security fields,such as event monitoring, information forecasting, and public opinion monitoring. With development of social media, social networks such as Twitter, Facebook and Sina Weibo become one of the most important channels for information generation and dissemination, containing a large number of images and texts with emotional colors. Since data in social networks is important of understanding publics views and positions,which is widely used as a source of data for public opinion analysis. Many existing methods extract features directly from images as additional information for multimodal analysis, which tends to neglect highdimensional semantic information, such as expressions, actions, etc., that exist locally in the image. To solve these problems, this paper proposes a feature extraction framework that combines local semantic information and a multimodal sentiment analysis method that combines visual and textual features. We use image description method to extract image features, extract text features by multilayer convolution, and finally train the classifier to combine these two features for decision making. The experimental results on the public data set MAVA in the field of sentiment analysis show that the model can effectively capture the graphic features and have better performance in the sentiment analysis task.

Key words: public opinion analysis, sentiment analysis, feature extraction, image description, multimodal, neural network