Estimating Divergence-or-Convergence in Discussions based on Abstractness of Words in Utterance

Abstract

Estimating divergence-or-convergence of discussions is expected to increase the productivity of the discussions. Divergence means expanding ideas and convergence means summarizing ideas. However, previous estimating methods can only be applied to limited discussions. In this study, divergence and convergence are estimated using the abstractness of utterances. Divergence is assumed to be a concrete state and convergence an abstract state, respectively. The estimated results in the three discussions are compared to human judgment. As a result, agreement rates are 69.2%, 87.9% and 54.3%. Additionally, the abstractness of utterances in two discussions are low in divergence and high in convergence. The activeness in three discussions may cause the difference of the abstractness in utterances. These results suggest that abstractness of utterances has a potential to estimate divergence-or-convergence in general discussions.