Table 4.

Summary of the protein-level prediction

Text Type	Language Model	Classifiers
		LR	RF	SVM	GBM
Titles	TFIDF	0.7703	0.7474	0.8330	0.6901
	LDA	0.5654	0.5723	0.5836	0.6227
	DEEP	0.6651	0.6698	0.7557	0.6826
	PDEEP	0.6611	0.5278	0.4314	0.6021
Abstracts	TFIDF	0.8132	0.8225	0.8208	0.7833
	LDA	0.5459	0.5739	0.5342	0.5713
	DEEP	0.7650	0.8105	0.7747	0.7909
	PDEEP	–	–	–	–
Function Descriptions	TFIDF	0.7412	0.7439	0.7715	0.6947
	LDA	0.6128	0.6829	0.6582	0.7065
	DEEP	0.8929	0.8962	0.9184	0.8788
	PDEEP	0.7017	0.7211	0.3474	0.6917

F-score was reported. The values shown are the average of the test sets in the five-fold cross validation. LR, Logistic Regression; RF, Random Forest; SVM, Support Vector Machine; GBM, Gradient Boosted Machine. For each text type, titles, abstracts and function descriptions, the best performing language model under four classifiers is highlighted in bold.