Skip to main content

Table 13 Performance of models learned from the CHEMDNER training set when evaluated on the development set.

From: Optimising chemical named entity recognition with pre-processing analytics, knowledge-rich features and heuristics

  Macro Micro
  P R F1 P R F1
Default features 86.66 79.01 80.89 88.55 76.82 82.27
Enriched features 88.26 81.11 82.86 89.87 78.99 84.07
Margin +1.6 +2.1 +1.97 +1.32 +2.17 +1.8