Skip to main content

Table 6 Impact of co-reference resolution on relation detection

From: The contribution of co-reference resolution to supervised relation detection between bacteria and biotopes entities

  Development Test
  R P F R P F
T2: none 0.582 0.338 0.428 0.608 0.324 0.423
T2: sim 0.583 0.433 0.497 0.598 0.422 0.495
T2: ana 0.559 0.307 0.397 0.602 0.311 0.410
T2: all 0.595 0.411 0.486 0.587 0.389 0.468
T3: none 0.368 0.312 0.338 0.360 0.281 0.316
T3: sim 0.388 0.399 0.394 0.347 0.356 0.351
T3: ana 0.317 0.273 0.293 0.334 0.279 0.304
T3: all 0.391 0.367 0.379 0.333 0.357 0.345
  1. Impact of co-reference resolution on relation detection. These experiments use the optimal (i.e., full '+context') set of features tested in Table 7, hence the best results here are optimal with respect to feature sets too. none = no co-reference resolution; sim = string similarity; ana = anaphora; all = sim+ana. Boldface highlights the optimal value on the development corpus, italics the associated value on the test corpus