Supervision of Webcasting-Anchor Behavior Evaluation Based on Barrage Emotion Analysis

摘要

With the rapid development of mobile computing, webcasts have been a novel form of entertainment. However, it is difficult to monitor anchors' behavior for live streaming. The anchors maybe insults viewers or uses pornography to seduce viewers. "Barrage" as a webcast commentary carrier can enrich the interactivity. We find that the content of the barrages can truly reflect the situation of an anchor. For this, we put forward a novel barrage-based anchor behavior evaluation system employing supervisor learning technology. We evaluate the emotion of the anchor based on the content of barrages, and predict the follow-up behavior of the anchor. If the content of the studio is violent or pornographic, the platform manager can ban him in advance to avoid even greater damage. The extensive experiment results show that the precision and recall of negative emotions can be up to 71.4% and 72.6%, respectively.