Skip to main content

Table 2 Performance comparing pyMeSHSim, DNorm, TaggerOne to Nelson’s manual work with similarity threshold set to 1

From: pyMeSHSim: an integrative python package for biomedical named entity recognition, normalization, and comparison of MeSH terms

MethodRecallaPrecisionbF1c
pyMeSHSim (with SCRs)0.940.560.70
pyMeSHSim (no SCRs)0.940.540.68
DNorm0.320.620.42
TaggerOne0.490.640.55
  1. a\( all=\frac{TP}{TP+ FN} \), where TP (true positive) is the number of phenotypes whose parsing results matched the manual work at determined similarity threshold. The similarity between MeSH terms identified by the two methods were measured with Lin score, and called as a TP or FP when their similarity was higher or lower than the determined threshold. FN (false negative) is the number of unrecognized phenotypes.
  2. b\( cision=\frac{TP}{TP+ FP} \), where FP is the number of phenotypes whose parsing results mismatched the manual work at determined similarity threshold.
  3. c\( 1=\frac{2\times precision\times recall}{precision+ recall} \) .