The statistics of test.jsonl can't match those in the paper.
XtHhua opened this issue · comments
I counted the number of mentions in the test set, and the results showed that 857 documents contained 41,486 mentions, which did not match the 20,557 mentions reported in the paper.
I regarded each dict in the "event_mentions" list of each document in the test set as a mention. Then the statistical result was 41486.