Agenda Extraction from Assembly Minutes by exploiting Relation between Questions and Answers
Ryoto Ohsugi, Tomoyosi Akiba, Shigeru Masuyama · 2021
Obtaining agenda discussed in assembly minutes is useful for grasping their structure. In this paper, we propose a method to automatically extract a list of agenda items from given minutes. Different from the well-known topic detection, the agenda extraction is required to output the short phrases that directly express agenda items without duplication. We cast the problem as extracting the linguistic expressions directly from the minutes and clustering them to remove their duplication. We propose different clustering methods for this purpose focusing on the relationship between questions and answers in the minutes. To see the effectiveness of our methods, we also conduct the experimental evaluation by comparing the agenda items obtained by our methods with two kinds of human created agenda.