Loading...

Extractive Meeting Summarization through Discourse Analysis

Bokaei, Mohammad Hadi | 2016

753 Viewed
  1. Type of Document: Ph.D. Dissertation
  2. Language: Farsi
  3. Document No: 48206 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Sameti, Hossein
  7. Abstract:
  8. Improvement of automatic speech recognition systems and the growth of audio data (such as broadcast news, voice mail, telephony conversations and meetings) have attracted plenty of research interest in the field of speech summarization. The goal of this dissertation is to improve the performance of the speech summarization in the domain of multi-party conversations, specifically meetings. Most of the previous work in this field are inheritted from the text summarization counterpart, whithout paying much attention to the discourse specific information of the multi-party conversations. The main idea of this work is to use discourse information to improve the accuracy of extracted summaries in meeting domain. We investigate the role of discourse segmentation of a conversation, as a level of discourse analysis, on the performance of meeting summarization algorithms. We propose a new criterion to segment a meeting transcript, namely function segmentation, in which the discourse is segmented accrording to the functional roles of the participants. After that, specific unsupervised algorithms are proposed to segment a meeting transcript into functionally coherent parts. These algorithms are evaluated on a manually labeled test set and compared with previous related work. Results show a significant improvement of the proposed algorithm over previous ones. We then investigate the influence of this segmentation on summarization task. We show that using the information in function segmentation will significantly improve the perfomance of the summarization algorithms. Evaluations on the standard AMI corpus prove the superiority of our proposed summarization algorithm over other baseline and state-of-the-art algorithms according to various standard metrics (classification p/r/f and ROUGE)
  9. Keywords:
  10. Unsupervised Learning ; Summarization ; Extractive Summarization ; Extractive Meeting Summarization ; Functional Segmentation ; Discourse Analysis ; Summary Keyword Extraction

 Digital Object List

 Bookmark

No TOC