Skip to main content
. Author manuscript; available in PMC: 2017 Jan 9.
Published in final edited form as: Artif Intell Med. 2015 Oct 3;66:29–39. doi: 10.1016/j.artmed.2015.09.007

Table 5.

Results of 10-fold cross-validation for different named-entity recognition methods.

Method Concept Precision (%) Recall (%) F1 score (%) TP FP TN
Dictionary-based Anatomy 77.7 53.2 63.1 2345 672 2066
Modifier 63.3 46.6 53.7 2349 1363 2687
Observation 72.7 49.0 58.5 1697 639 1769
Uncertainty 90.1 38.9 54.3 566 62 889
Total 71.8 48.4 57.8 6957 2736 7411
CMM Anatomy 90.5 92.6 91.5 4084 429 327
Anatomy modifier 80.2 74.2 77.1 1363 336 474
Observation 89.4 84.9 87.1 2941 350 525
Observation modifier 81.4 79.5 80.4 2543 581 656
Uncertainty 91.1 78.5 84.3 1142 111 313
Total 87.0 84.0 85.5 12,073 1807 2295
CRF Anatomy 90.6 92.6 91.6 4085 425 326
Anatomy modifier 80.8 74.4 77.5 1366 324 471
Observation 89.4 84.1 86.7 2915 344 551
Observation modifier 81.2 79.5 80.3 2542 587 657
Uncertainty 90.7 76.0 82.7 1105 114 350
Total 87.0 83.6 85.3 12,013 1794 2355