Semantic Similarity based on Weighted Ontology - Geographical Information Systems: Trends and Technologies - page 320

Table 6. Results about similarity methods by changing occurrences of attributes.

Correlation

Att. Lin Dice MDSM v MDSM c GSim v GSim c

15 0.7127 -0.1539 0.1495 0.2290 0.3335 0.7191

17 0.7127 -0.1021 0.1836 0.2361 0.3902 0.7248

19 0.7127 0.0332 0.2570 0.2640 0.5708 0.7751

21 0.7127 0.1337 0.2984 0.2865 0.6067 0.7772

23 0.7127 0.1855 0.3351 0.3103 0.6503 0.7903

25 0.7127 0.2972 0.3774 0.3419 0.7214 0.8198

27 0.7127 0.5034 0.5041 0.4532 0.7748 0.8310

29 0.7127 0.5705 0.5490 0.5052 0.7997 0.8340

31 0.7127 0.5840 0.5422 0.5125 0.8148 0.8342

33 0.7127 0.6132 0.5413 0.5260 0.8242 0.8343

35 0.7127 0.6135 0.5446 0.5430 0.8332 0.8345

1

Lin
Dice
MDSM v
MDSM c
GSim v
GSim c

0.8

0.6

0.4

0.2

0

-0.2

15 19 23 27 31 35

Attributes

Fig. 10. Correlation of methods by varying occurrence of attributes.

Color image of this figure appears in the color plate section at the end of the topic.

MDSM v , MDSM c . Considering that both Dice and MSDM are feature-based
approaches, the Dice method achieves better correlation than MDSM by
increasing the number of shared features. However, their correlations are
lower than the correlations obtained by GSim method. Since Lin method