Identify group roles by text mining on group discussion in a Web-based learning system

To explore how group member behaviour influences group performance, most studies stress on the influences of group leadership. However, the existence of member-roles is also an effective indicator of group performance, meanwhile; teachers have difficult to extract the member-roles which students play on a Web-based learning system. This research (1) assist teachers to extract group social communication patterns when monitoring students´ social communication (2) assist teachers to extract member-roles that may influence the group learning performance, and (3) assist teachers to extract the influence of member-roles on group learning performance and help teachers to promote group to learn better on Web group learning environment. This research derives 11 member-roles by exploring communication with teachers expert knowledge and criteria, and extracting the influence on four primary group performance indicators: (1) individual grades, (2) group project grade, (3) resource-sharing frequency, and (4) group drop out rate. The statistical analysis method shows that most of the member-roles have significant influence on group learning performance.