Semisupervised Learning Based Disease-Symptom and Symptom-Therapeutic Substance Relation Extraction from Biomedical Literature

Table 3

The results obtained with Co-Training on the disease-symptom test set. Combination method integrates three classifiers (the feature kernel, graph kernel, and tree kernel) with the same weight.

Combination	View			-score	AUC
Feature and graph kernel	Feature kernel	88.32	67.97	76.82	88.01
	Graph kernel	83.26	71.88	77.15	87.54
	Combination	74.91	85.16	79.71	88.66
Feature and tree kernel	Feature kernel	86.06	69.92	77.15	88.51
	Tree kernel	57.80	92.58	71.17	74.99
	Combination	75.08	87.11	80.65	87.18
Graph and tree kernel	Graph kernel	84.04	69.92	76.33	86.04
	Tree kernel	58.10	95.31	72.19	78.10
	Combination	82.43	76.95	79.60	86.84