Average Mention Representation

AbstractThe distribution of knowledge elements such as entity types and event types is long-tailed in natural language. Hence information extraction datasets naturally conform long-tailed distribution. Although imbalanced datasets can teach …...