Table 4.
Influences of curated CDR articles on the relation extraction results
| Method | Intra-sentence level | Inter-sentence level | Document level | ||||||
|---|---|---|---|---|---|---|---|---|---|
| P (%) | R (%) | F (%) | P (%) | R (%) | F (%) | P (%) | R (%) | F (%) | |
| KCN | 70.61 | 60.41 | 65.12 | 65.37 | 12.57 | 21.09 | 69.65 | 72.98 | 71.28 |
| -train&test | 64.44 | 53.38 | 58.39 | 45.41 | 8.35 | 14.10 | 60.98 | 61.73 | 61.35 |
| -train | 64.07 | 60.23 | 62.09 | 54.63 | 11.07 | 18.41 | 62.40 | 71.29 | 66.55 |
| -test | 65.53 | 48.87 | 55.99 | 49.71 | 8.16 | 14.02 | 61.76 | 57.41 | 59.50 |
| Only KB | 59.44 | 65.85 | 62.48 | 31.34 | 21.49 | 26.36 | 50.41 | 87.24 | 63.90 |
The descriptions and analysis for Table 4 could be found in subsection “Influences of the curated articles in the CDR dataset”. The highest scores are highlighted in bold