A coherent graph-based semantic clustering and summarization approach for biomedical literature and a new summarization evaluation method

Table 4 Summary of Overall Experiment Results on MEDLINE Document Sets

	STC		K-means	Original Bisecting K-means [25]	CLUTO Bisecting K-means		CSUGAR
	word strings	concept strings			Largest	LOS
MI	μ: 0.429 σ: 0.238	μ: 0.359 σ: 0.149	μ: 0.128 σ: 0.148	μ: 0.395 σ: 0.193	μ: 0.161 σ: 0.139	μ: 0.096 σ: 0.112	μ : 0.053 σ : 0.031
Purity	μ: 0.601 σ: 0.214	μ: 0.731 σ: 0.098	μ: 0.932 σ: 0.080	μ: 0.666 σ: 0.154	μ: 0.918 σ: 0.064	μ: 0.944 σ: 0.056	μ : 0.947 σ : 0.030
F-measure	μ: 0.499 σ: 0.285	μ: 0.512 σ: 0.198	μ: 0.828 σ: 0.206	μ: 0.532 σ: 0.236	μ: 0.780 σ: 0.180	μ: 0.880 σ: 0.139	μ : 0.926 σ : 0.062

LOS: selecting the cluster (to be bisected) with the least overall similarity and Largest: selecting the largest cluster to be bisected. MI: the smaller, the better clustering quality. Purity and F-measure: the bigger, the better clustering quality.

ISSN: 1471-2105