SENTIMENT ANALYSIS OF AUDIO, VISUAL AND TEXTUAL DATA

This review article introduces a new method for sentiment analysis on an example of news videos based on a combination of text, audio and visual information. The method aimed at conducting research using modern methods of emotional recognition using facial expressions, audio diagnostics and video sentiment analysis. Experimental results show that the approach achieves accuracy up to 84% in the tasks of sentiment classification (tension levels). This allows us to say that it is very promising for multimedia analytics in the journalistic sphere and not only.