|
References | Methods | Contributions | Limitations |
|
Ayush et al. [44] | Developed integral model including probability and fuzzy models for determination of human constitutional types | Proposed MDES system creates and supports decision system to users via providing reliable information about disease manifestation | It has not enough practical evidence for effectiveness and efficiency |
Korenevskiy [45] | Synthesis of fuzzy decision rules | Simple to calculate with high possibility of diagnosis and predetermined level of reliability | It requires larger training samples |
Atutxa et al. [21] | ICD-10 encoding based on neural networks | Multilingual ICD-10 coding. The method is interpretable and it outperforms alternative approaches. | Worse performance was detected on larger datasets |
Combi et al. [22] | MagiCoder, an NLP algorithm | Simple, efficient in terms of computational complexity for Italian pharmacovigilance language | Inability to handle negations in textual medical records |
Lu et al. [14] | Combined classic enhanced sequential inference model (ESIM) and BiLSTM network | Achieved higher accuracy compared to existing methods without knowledge enhanced | Challenges of concepts with multiple definition was not addressed |
Kloehn et al. [24] | Proposed a novel algorithm SubSimplify | Improved quality in English and Spanish by providing multiword explanations for difficult terms | There is a possibility of the proposed model generating incomplete explanations |
Sarker et al. [25] | Combination of fuzzy matching and intersection | Increased accuracy against human annotations | Inability to detect negations in expressions |
|