Figure 3.

Performance comparison of different encoding methods. For deep learning models with different encoding methods, the area under the curves (AUCs) in the independent tests were averaged over all species and their variances were calculated. W2V, C-BE, C-NC, BE, NC, BE+NC and C-BE+C-NC correspond to word2vec, contextual-binary encoding, contextual-NCPNF, binary encoding, NCPNF, combination of binary encoding and NCPNF and combination of contextual-binary encoding and contextual-NCPNF, respectively. One-sided, paired-sample t-tests were conducted to compare the AUCs between MNBE and C-MNBE and the AUCs between NCPNF and C-NCPNF.