视频学习资源中关键词的抽取方法研究

Research on Keyword Extraction Method of Video Learning Resources

近年来知识图谱广泛应用于在线智能学习系统,在知识图谱的自动构建过程中,关键词抽取是其首要环节。在当前众多关键词抽取方法中TextRank算法应用最为广泛。在线学习系统有大量视频学习资源,而目前的关键词抽取方法主要用来处理论文、新闻报道等文本资源。为从视频学习资源中抽取关键词,在TextRank算法基础上,提出一种改进的关键词抽取方法TW-TextRank。首先通过语音识别把视频资源转换为文本,然后获取关键词在视频中的时长并计算权重,最后进行关键词抽取和排序。实验以8门C语言教学视频作为素材,和传统TextRank算法进行比较,结果表明TW-TextRank算法对召回率和F值有较为明显的提升效果,在视频资源的知识图谱构建方面有较强实用性。

Knowledge Graph is widely used in online intelligent learning system in recent years.Keyword extraction is primary link in the process of automatic construction of Knowledge Graph.TextRank algorithm is the most widely used in many keyword extraction methods.Online learning system has a large number of video learning resources,and the current keyword extraction methods are mainly used to deal with text resources such as papers and news reports.In order to extract keywords from video learning resources,an improved keyword extraction method TW-TextRank is proposed based on TextRank algorithm.Firstly,the video resources are converted into text through speech recognition,then the duration of keywords in the video is obtained and the weight is calculated,and finally the keywords are extracted and sorted.The experiment takes 8 C programming teaching videos as materials and compares them with the traditional TextRank algorithm.The results show that TW-TextRank algorithm has obvious improvement effect on the Recall rate and F value,and has strong practicability in the construction of knowledge map of video resources.