Skip to main content

Table 3 Performance measurement of different system settings on the GRN training data.

From: Application of the EVEX resource to event extraction and network construction: Shared Task entry and result analysis

Method Dataset SER F Rel. P Rel. R Rel. F Rel. SER
TC All EVEX data 1.56 8.86 39.29% 41.98% 40.59% 1.23
TC, ETF All EVEX data 1.15 11.53 59.74% 35.11% 44.23% 0.89
TC, ETF B. subtilis PMIDs 0.954 20.81 71.43% 22.90% 34.68% 0.86
TC, ETF GRN PMIDs 0.939 17.39 80.00% 18.32% 29.81% 0.86
  1. The SER score is the main evaluation criterion of the GRN challenge. The relaxed precision, recall, F and SER scores are produced by scoring the predictions regardless of the specific event types. TC refers to rule-based EVEX-GRN type conversion, ETF to rule-based entity type filtering. The ML-based methods are not shown because they are trained on this dataset.