Skip to main content

Table 7 Performance by corpus size.

From: Mining clinical relationships from patient narratives

   Corpus size
Relation Metric C25 C50 C77
has_finding Count 91 216 311
  P 66 63 65
  R 74 74 81
  F1 67 67 72
has_indication Count 91 117 224
  P 22 25 42
  R 30 31 47
  F1 23 25 42
has_location Count 127 199 364
  P 72 71 72
  R 76 80 81
  F1 73 74 75
has_target Count 51 90 136
  P 65 49 62
  R 60 65 66
  F1 59 54 63
laterality_modifies Count 57 73 128
  P 77 78 84
  R 69 68 88
  F1 72 69 84
negation_modifies Count 34 67 101
  P 78 79 79
  R 80 93 93
  F1 78 84 85
sub_location_modifies Count 30 43 76
  P 64 91 88
  R 64 85 95
  F1 64 86 91
Overall Count 481 805 1340
  P 62 63 65
  R 65 71 75
  F1 63 66 69
  1. Variation in performance by training corpus size. The "Count" row gives the number of training instances of a relation type, for the given corpus. The cumulative feature set +event of Table 4 was used.
\