Large-Scale Automatic Labeling of Video Events with Verbs Based on Event-Participant Interaction