Table 4 NLP performance in training and test dataset
From: Initial development of tools to identify child abuse and neglect in pediatric primary care
Resource | Sample | NLP identified CAN | Manual review identified CAN | Precision |
|---|---|---|---|---|
Training Document | 14,397 | 422 | 396 | 93.8% |
Training Sentence | 1,486 | 1,349 | 93.8% | |
Test Document | 4,854 | 105 | 78 | 74.2% |
Test Sentence | 311 | 222 | 71.4% |