Research on Highlight Snippets Identification Technology Based on Sentiment Analysis of Bullet Curtain

At present, the method of extracting highlights snippets of video is restricted to image and audio analysis in sports videos. About this problem, this paper proposes a new method of using bullet curtain text as sentiment carrier for sentiment analysis, and then using its time coupling to identify the highlight snippets. Firstly, according to the particularity of bullet curtain text, a domain sentiment dictionary is constructed. Secondly, the paper proposes a new feature integrating time series of bullet curtain, and classifies bullet curtain text by support vector machine. Finally, bullet curtain with intense sentiment is superimposed on the video time scale, and the full video brilliance curve is obtained. The best threshold is selected to identify highlight snippets. Experimental results show that the proposed method has higher accuracy, precision and recall rate, and the validity of the proposed method is verified.

I. Introduction

With the rapid spread of the Internet, related technologies such as multimedia processing technology, network transmission, and video data processing are rapidly developing, and video data from different fields is growing at an amazing rate. For example, YouTube has 4.67 billion video playbacks per month in total. Hulu, another professional video site, also has more than 200,000 high-quality videos whose monthly visits have remained at 900 million. Compared to watching a complete video, sometimes users prefer to watch some of these video snippets, such as highlights snippets of a humorous movie or an exciting collection of soccer goals. Subscribers even want to search the video snippets that are similar to the emotions based on videos they have seen before. The major video sites also began to manually tag highlight nodes on the video timeline.