Skip to main content

Advertisement

Table 9 F-Measure results using all the features and all but one of the features

From: A comparison and user-based evaluation of models of textual information structure in the context of cancer risk assessment

   ALL A B C D E F G H I J K
S1 OBJ .90 .89 .87 .92 .90 .90 .91 .91 .91 .92 .91 .88
  METH .80 .81 .80 .80 .79 .81 .79 .80 .80 .80 .81 .81
  RES .88 .90 .88 .90 .88 .90 .88 .88 .88 .89 .89 .90
  CON .86 .85 .82 .87 .88 .90 .90 .88 .89 .88 .88 .90
S2 BKG .91 .94 .90 .90 .93 .94 .94 .91 .93 .94 .92 .94
  OBJ .72 .78 .84 .78 .83 .88 .84 .81 .83 .84 .78 .83
  METH .81 .83 .80 .81 .80 .85 .80 .78 .81 .81 .82 .83
  RES .88 .90 .88 .89 .88 .91 .89 .89 .90 .90 .90 .89
  CON .84 .83 .77 .83 .86 .88 .86 .87 .88 .89 .88 .81
  REL - - - - - - - - - - - -
  FUT - 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0
S3 HYP - - - - - - - - - - - -
  MOT .82 .84 .80 .76 .82 .82 .83 .78 .83 .83 .83 .83
  BKG .59 .60 .60 .54 .67 .62 .62 .59 .61 .61 .62 .61
  GOAL .62 .67 .67 .62 .71 .62 .67 .43 .67 .67 .67 .62
  OBJT .88 .85 .83 .74 .83 .85 .83 .74 .83 .83 .83 .85
  EXP .72 .68 .72 .53 .65 .70 .72 .73 .74 .74 .72 .68
  MOD - - - - - - - - - - - -
  METH .87 .86 .87 .66 .85 .89 .87 .88 .86 .86 .87 .86
  OBS .82 .81 .84 .72 .80 .82 .81 .80 .82 .82 .81 .81
  RES .87 .87 .88 .74 .87 .86 .87 .86 .87 .87 .87 .88
  CON .88 .88 .82 .88 .83 .87 .87 .84 .87 .88 .87 .86
  1. A-K: History, Location, Word, Bi-gram, Verb, Verb Class, POS, GR, Subj, Obj, Voice
  2. We have 1.0 for FUT in S2 probably because the size of the training data is just right, and the model doesn't over-fit the data. We make this assumption because we have 1.0 for almost all the categories on the training data, but only for FUT on the test data.