Table 4
Precision of SecTag Component Methods to Identify Sections
| Method | Count (%) | Number Correct | Precision (95% CI) |
| Labeled Sections | |||
| Exact or normalized match | 11221 (70.0%) | 11123 | 99% (98.9–99.3) |
| Variant generation | 130 (0.8%) | 110 | 85% (77–90) |
| Unlabeled sections | |||
| Bayesian prediction | 1867 (11.6%) | 1503 | 81% (79–82) |
| Next-section rules | 29 (0.2%) | 27 | 93% (77–92) |
| NLP | 2112 (13.2%) | 1939 | 92% (91–93) |
| Both labeled and unlabeled sections | |||
| Spelling correction | 53 (0.3%) | 33 | 62% (48–75) |
| Labels within a sentence | 471 (2.9%) | 444 | 94% (92–96) |
| Modifier removal | 153 (1.0%) | 150 | 98% (94–100) |
| Totals | 16036 | 15329 | 96% (95, 96) |
-
CI = confidence interval; NLP = natural language processing.
