摘要
随着互联网时代的到来,移动视频市场迅速发展,相关媒体技术应运而生,从而出现了弹幕这种新兴的视频评论方式,并在许多在线视频网站得到应用。弹幕是典型的短文本,能够允许用户在观看视频时实时评论视频内容和表达自身情感,具有刺激用户互动、帮助后续观看者理解视频的功能。弹幕评论和传统视频评论共同覆盖了用户观看视频前后的全过程情感,对二者进行情感分析可以对视频内容质量进行相对全面的预估,能够弥补视频网站缺乏用户对视频评分的空缺。因此,本文提出一种基于双层情感分析的视频推荐算法,首先借助隐含狄利克雷分布(Latent Dirichlet Allocation,LDA)主题模型筛选用户偏好的视频主题,其次通过对视频评论进行情感极性分析来对视频打分,最后利用视频弹幕计算待推荐视频与用户历史视频的情感相似度,生成视频推荐列表。实验表明,该方法能够充分利用弹幕和评论的情感信息,提升了视频推荐的准确率。
With the advent of the Internet era, the mobile video market has developed rapidly, and relevant media technologies came into being. As a result, bullet screen, a new way of video comment, has been applied in many online video websites. Bullet screen is a typical short text, which can allow users to comment on the video content and express their emotions in real time when watching the video. It has the function of stimulating user interaction and helping subsequent viewers understand the video. Bullet screen reviews and traditional video reviews cover the whole process of emotion before and after users watch the video. Emotion analysis of them can make a relatively comprehensive prediction of the quality of video content, and can make up for the lack of users’ video rating on video websites. Therefore, this paper proposes a video recommendation algorithm based on double-layer emotion analysis. Firstly, the video topics preferred by users are selected by using the Latent Dirichlet Allocation(LDA) topic model, and then the video is scored by analyzing the emotional polarity of video comments, Finally, the emotional similarity between the video to be recommended and the user’s history video is calculated by using the video barrage to generate the video recommendation list.Experiments show that this method can make full use of the emotional information of barrage and comments, and improve the accuracy of video recommendation.
作者
张晟哲
ZHANG Shengzhe(Harbin Institute of Technology,Harbin Heilongjiang 150000,China)
出处
《信息与电脑》
2022年第6期74-77,共4页
Information & Computer
关键词
情感分析
弹幕文本
文本相似度
LDA主题模型
视频推荐算法
sentiment analysis
barrage text
text similarity
LDA topic model
the algorithm for video recommendation
