Semantic Similarity based on Weighted Ontology - Geographical Information Systems: Trends and Technologies - page 321

is independent from features, its correlation remains invariant by varying
occurrences of attributes.
In Table 7 and in Fig. 11, the correlations of the above mentioned
methods with human judgment by changing the occurrences of parts are
shown. We observe that the values obtained by the GSim method are higher

Table 7. Results about similarity methods by changing occurrences of parts.

Correlation

Parts Lin Dice MDSM v MDSM c GSim v GSim c

7 0.7127 0.2972 0.3552 0.4007 0.7928 0.6056

9 0.7127 0.2972 0.3585 0.3912 0.7839 0.6652

11 0.7127 0.2972 0.3621 0.3799 0.7737 0.7137

13 0.7127 0.2972 0.3659 0.3692 0.7621 0.7517

15 0.7127 0.2972 0.3697 0.3593 0.7499 0.7816

17 0.7127 0.2972 0.3737 0.3507 0.7364 0.8037

19 0.7127 0.2972 0.3774 0.3419 0.7214 0.8198

21 0.7127 0.2972 0.3821 0.3368 0.7050 0.8312

23 0.7127 0.2972 0.3864 0.3309 0.6873 0.8392

25 0.7127 0.2972 0.3908 0.3260 0.6672 0.8440

27 0.7127 0.2972 0.3948 0.3216 0.6460 0.8469

29 0.7127 0.2972 0.3984 0.3176 0.6237 0.8482

31 0.7127 0.2972 0.4014 0.3142 0.5996 0.8483

1

Lin
Dice
MDSM v
MDSM c
GSim v
GSim c

0.8

0.6

0.4

0.2

0

-0.2

7 11 15 19 23 27 31

Attributes

Fig. 11. Correlation of methods by varying occurrence of parts.
Color image of this figure appears in the color plate section at the end of the topic.