Open RicardoL1u opened 3 years ago
I have the same question. Did you try some experiments?
The maven dataset is mainly used for trigger words detection and event type classification, the mentions in the same event type class are not discussing the same event. So is it OK to use this dataset? Meanwhile, it seems there is no time/user info in the MAVEN dataset, so how to build a heterogeneous social graph in this case?
The maven dataset is mainly used for trigger words detection and event type classification, the mentions in the same event type class are not discussing the same event. So is it OK to use this dataset? Meanwhile, it seems there is no time/user info in the MAVEN dataset, so how to build a heterogeneous social graph in this case?