Table 4 NLP type and Algorithm performance characteristics of the included articles
Author | NLP type (algorithm) | Performance |
|---|---|---|
Hammami et al. [36] | Developed by researchers (rule-based) | 86.3% < = Sens < = 99.2% |
85.9% < = Prec < = 99.2% | ||
84.9% < = F1-score < = 99.2% | ||
98.10% < = Acc < = 99.2% | ||
Ryu et al. [37] | Developed by researchers (rule-based) | Sens = 100% |
98.6 < = Prec < = 100 | ||
99.3 < = F1-score < = 100 | ||
Oliveira et al. [41] | Developed by researchers Developed by researchers (rule-based + ML) + CLAMP software | – |
Becker et al. [48] | Developed by researchers (semi-automated rule-based system) | Sens = 99.54% |
Prec = 97.95% | ||
Wang et al. [40] | Developed by researchers (algorithms to generate final normalized concept names for each data element) + ML + Premade (Med Tagger) | 0.982 < = Recall < = 1 |
0.885 < = Prec < = 1 | ||
Kumar et al. [33] | Developed by researchers (semantic similarity measures and clustering methods) + cTAKES | P value < = 0.05 |
Wadia et al. [43] | Premade (cTAKES) | Sens = 77.3 |
Spec = 72.5 | ||
Prec = 88.4 | ||
NPV = 54 | ||
Bustos et al. [38] | Developed by researchers | 0.79 < = Sens < = 0.93 |
((Fast Text. CNN, SVM, KNN)) | 0.79 < = Prec < = 0.93 | |
0.79 < = F1-score < = 0.93 | ||
Faina Linkov et al.[47] | Premade (TIES) | – |
Sada et al. [34] | Premade (ARC = automated retrieval console) | 0.94 < = Sens < = 0.96 |
0.68 < = Spec < = 0.97 | ||
0.75 < = PPV < = 0.96 | ||
Nguyen et al. [44] | Premade | Sens = 0.78 |
( Medtex) + Developed by researchers (rule-based) | Prec = 0.83 | |
F1-score = 0.80 | ||
Hoogendoorna et al. [45] | Developed by researchers (rule-based) | 0.870 < = Accu < = 0.831 |
Löpprich et al. [39] | Developed by researchers (SVM—MEC) + German framed clinical text | 0.89 < = F1-score < = 0.92 |
Mehrabi et al. [32] | Developed by researchers (rule-based) | 87.8 < = Prec < = 88.1 |
Sippo et al. [42] | Premade (BROK) | Sens = 100% |
96.6% < = Prec < = 100% | ||
Segagni et al. [31] | Developed by researchers and premade | – |
(rule-based system for the onco-i2b2 + researcher-made algorithm | ||
Strauss et al. [46] | Premade (CoPathPlus) + Developed by researchers (Scent = rule-based) | Breast cancer |
0.74 > Sens > 1 | ||
Spec = 0.99 | ||
0. > 90 Prec > 0.94 | ||
0.97 < NPV > 1 | ||
Prostate cancer | ||
0.71 > Sens > 0.97 | ||
0.9 > 8 Spec < 0.99 | ||
0.88 > Prec > 0.97 | ||
0.95 > NPV > 0.99 |