Skip to main content
. 2022 Jul 6;23:263. doi: 10.1186/s12859-022-04810-y

Table 5.

Entity-level results of BiLSTM models.

Label Precision Recall F1 Support
Random initialization
DISEASE 0.4387 0.2913 0.3502 454
RAREDISEASE 0.4592 0.4712 0.4651 1095
SIGN 0.3288 0.3224 0.3256 958
SYMPTOM 0.0000 0.0000 0.0000 54
Micro-avg 0.3668 0.3742 0.3705 2561
Macro-avg 0.2454 0.2170 0.2282 2561
Macro-weighted 0.3946 0.3742 0.3820 2561
Google news
DISEASE 0.4432 0.3071 0.3628 454
RAREDISEASE 0.4796 0.4971 0.4882 1095
SIGN 0.3166 0.3419 0.3287 958
SYMPTOM 0.4571 0.3200 0.3765 54
Micro-avg 0.3724 0.4020 0.3866 2561
Macro-avg 0.3393 0.2932 0.3112 2561
Macro-weighted 0.4084 0.4020 0.4028 2561
Glove
DISEASE 0.4246 0.3622 0.3909 454
RAREDISEASE 0.5194 0.5529 0.5356 1095
SIGN 0.3114 0.3971 0.3491 958
SYMPTOM 0.6154 0.4800 0.5393 54
Micro-avg 0.3850 0.4596 0.4190 2561
Macro-avg 0.3742 0.3584 0.3630 2561
Macro-weighted 0.4236 0.4596 0.4387 2561
Wiki-pubmed-PMC
DISEASE 0.5794 0.4339 0.4962 454
RAREDISEASE 0.5378 0.5388 0.5383 1095
SIGN 0.3167 0.3570 0.3356 958
SYMPTOM 0.5946 0.4074 0.4835 54
Micro-avg 0.4170 0.4494 0.4326 2561
Macro-avg 0.4057 0.3474 0.3707 2561
Macro-weighted 0.4637 0.4494 0.4539 2561

Best micro and macro scores are in bold. Best scores for each entity type are also in bold