Detection the Role of a Participant in a Conversation by Auto-Moderator

Tai Wang, Xiangen Hu, and Chuang Gao


As an application of Semantic Representation Analysis (SRA), we implement Individualized Conversation Characteristics Curves (ICCC) as simple monitoring mechanism in typical online discussion forum where multiple discussants assume different roles. In this paper, we present some intuitive models for each of the roles and explore these models by simulations. The implementation is called “Auto-Moderator.” In our presentation, we will only consider four different roles in discussion forums: Leader, Follower, Loner, and Wonderer. Leaders are those discussants whose views/opinions are most influential such that others are following him/her in consequent discussions. Followers are those who only follow other’s opinions, Loners are those who ignore any previous views/opinions and insist on having their own views/opinions. Wonderers are those none of the other three. We build models for each of the roles by semantically analyzing the ICCCs for each of the discussants. This is one of the first papers for this study. We will 1) introduce the concept of ICCC, 2) Propose models for each of the roles, 3) present a methods of evaluating those models, and 4) present simulation results.