文章基本信息

标题：Discussion Summarization Based on Hierarchical Structure Using Verbal and Non-Verbal Information
本地全文：下载
作者：Hiroya Miura ; Yoshinari Takegawa ; Asuka Terai 等
期刊名称：Lecture Notes in Engineering and Computer Science
印刷版ISSN：2078-0958
电子版ISSN：2078-0966
出版年度：2018
卷号：2237&2238
页码：308-313
出版社：Newswood and International Association of Engineers
摘要：The aim of the study is to analyze the semantic relationships among utterances in a meeting at a deep level. To achieve the aim, we propose a novel tree structure, called discussion time-span tree, for representing a discussion structure based on the relative importance of utterances contained which are derived from the meaning of a meeting. A discussion time-span tree is a binary tree, each leaf of which corresponds to a utterance; it has the following characteristics: adjacent relevant utterances are hierarchically grouped, and a more important utterance becomes a head. Thus, a discussion timespan tree represents implicit meaning and intention contained in meeting records, and can be used for summarization. In the study, we propose a method and rules to generate a discussion time-span tree based on verbal and non-verbal information. In the evaluation on important utterance extraction using the proposed system, it was shown that the score of ROUGE-2 was higher than that of the existing document summarizing technique. Moreover, in the evaluation by subjects conducted by 10 subjects, we showed that the system can grasp conference contents efficiently, and demonstrated the usefulness of the system.
关键词：automatic summarization system; discussion analysis; information retrieval