BioMed Research International / 2015 / Article / Tab 5

Research Article

Recognition and Evaluation of Clinical Section Headings in Clinical Documents Using Token-Based Formulation with Conditional Random Fields

Table 5

Performance comparison among different methods.

DatasetConfiguration (%) (%) (%)

Set 2Dict. method 1 (SecTag)19.979.3131.82
Dict. method 1 (set 1)52.1894.0467.12
Dict. method 1 (SecTag + set 1)23.1994.9933.47
Dict. method 2 (SecTag)41.1979.3154.22
Dict. method 2 (set 1)75.594.0483.76
Dict. method 2 (SecTag + set 1)45.3394.9961.37
Sentence-based formulation (ME)81.5482.1681.85
Token-based formulation (CRF)95.4892.6694.05

TestDict. method 1 (SecTag)21.1580.2333.47
Dict. method 1 (set 1 + set 2)54.1394.8768.93
Dict. method 1 (SecTag + set 1 + set 2)24.3895.4838.84
Dict. method 2 (SecTag)41.7280.2354.89
Dict. method 2 (set 1 + set 2)76.3794.8484.6
Dict. method 2 (SecTag + set 1 + set 2)45.5995.4861.71
Sentence-based formulation (ME)85.4685.5485.5
Token-based formulation (CRF)96.0492.494.19