Skip to main content
. 2018 Jun 13;19(Suppl 8):209. doi: 10.1186/s12859-018-2195-1

Table 2.

Results obtained for max-pooling CNN on the test dataset without negative filtering

Classes P R F1
Advise 79.33% 64.25% 71.00%
Effect 68.90% 54.17% 60.65%
Int 81.08% 31.25% 45.11%
Mechanism 58.29% 70.57% 63.84%
Overall 67.13% 59.22% 62.93%