Skip to main content

Table 2 Results over test set

From: Detecting modification of biomedical events using a deep parsing approach

Mod RMRS from Extra R P F
SPEC - W + 3 - 3 6.97 23.73 10.77
SPEC ERG - 8.96 41.86 14.75
SPEC fb(ERG, RASP) - 12.44 52.08 20.08
SPEC cb(ERG, RASP) - 6.47 41.94 11.21
SPEC ERG W + 3 - 3 11.44 26.14 15.92
SPEC fb(ERG, RASP) W + 3 - 3 9.95 28.17 14.71 †
SPEC cb(ERG, RASP) W + 3 - 3 7.46 24.19 11.41
NEG - W + 3 - 3 19.55 30.94 23.96
NEG - Negex 17.83 12.73 14.85
NEG ERG - 11.82 35.62 17.75
NEG fb(ERG, RASP) - 13.64 34.09 19.48
NEG cb(ERG, RASP) - 12.73 33.33 18.42
NEG ERG W + 3 - 4 19.55 32.33 24.36
NEG fb(ERG, RASP) W + 3 - 4 19.55 32.58 24.43
NEG cb(ERG, RASP) W + 3 - 4 20.91 41.07 27.71
  1. Results over the test data using the UTurku Task 1 system ("fb" = fallback strategy, where we use the first source if possible, otherwise the second; "cb" = use undifferentiated RMRSs from each source to create feature vectors). † denotes the feature set which performed best over the development set using gold Task 1 annotations.