Table 6.

Macrofactors selected by the multilevel factor elimination (MFE) algorithm in the Bidirectional Encoder Representations from Transformers for Biomedical Text Mining (BioBERT) model across different datasets.

MFE algorithm	On the basis of the Revised JNLPBAa dataset	On the basis of the BC5CDRb dataset	On the basis of the AnatEMc dataset
Input	sLend, eLene, eNumf, eDeng, elConh, and tEWCi	sLen, eLen, eNum, eDen, elCon, and tEWC	sLen, eLen, eNum, eDen, elCon, and tEWC
Layer 1	eLen, eNum, elCon, and tEWC	eLen, eNum, eDen, and tEWC	sLen, eLen, eNum, and eDen
Layer 2	eLen and eNum	eNum and tEWC	eLen and eNum
Layer 3	eNum	eNum	eNum

^aJNLPBA: Joint Workshop on Natural Language Processing in Biomedicine and its Applications.

^bBC5CDR: BioCreative V CDR.

^cAnatEM: Anatomical Entity Mention.

^dsLen: sentence length.

^eeLen: entity phrase length.

^feNum: number of entity words in each entity phrase.

^geDen: entity density.

^helCon: entity label consistency.

ⁱtEWC: total entity word count in each entity type.