基于主题的SVM与MMR融合的会议摘要技术

Meeting summarization technology confused with SVM and MMR based on theme

为获得较高质量的会议摘要,提出一种基于主题的MMR与SVM相融合的会议摘要提取方法。以会议文本记录为处理对象,以SVM(support vector machine)和MMR(maximal marginal relevance)摘要提取算法为基础,针对当前的会议围绕某个主题进行讨论、交流的特点,以主题关键词为依据进行打分。实验结果表明,将所提系统分别与SVM摘要系统、MMR摘要系统以及SVM与MMR相结合摘要系统进行对比时,前者提取的摘要效果更好。

To obtain a high quality conference summary,a topic based MMR and SVM fusion method was proposed.The meeting text records was taken as the processing object,SVM（support vector machine）and MMR（marginal relevance maximal）algorithm were used to extract the basis.According to the characteristics of the current meeting,around a theme of communication,the theme of keywords was taken as the basis for scoring.Experimental results show comparing with the SVM summary system,MMR summary system and SVM and MMR combined summary system,the abstract extraction effects of the proposed system are better.