A Method for Online Discussion Design and Discussion Data Analysis

Abstract

The final goal of our research has been conducting research and development on a large-scale consensus support system in which we will introduce automated facilitators by applying deep learning technology. The aim of this paper is to present how to design discussions on the online discussion system which is not consisted by explicit framework and how to process those discussion data as training data of deep learning for the development of automated facilitation system. As a first step to attain a consensus formation, it is necessary to design discussions constructively. In order to do so, participants’ opinions must be collected efficiently. The issue-based information system (IBIS) is a well-known efficient way to do this. In a discussion adopting the IBIS idea, participants can understand each other’s opinions clearly and propose their new ideas smoothly. It is possible to annotate the word data collected from online discussions with the constituent elements of IBIS. The annotated data is reusable as training data of deep learning and intended for application to other systems as open data. Based on above, we conducted an online discussion design experimental method and examined online discussion with applying IBIS idea. Our experiments proved that it is possible to extract IBIS elements in non-framed online discussions.