Skip to main content

Table 5 Basic characteristics of the Google News word embedding-based similarity networks obtained for various slicing thresholds on cosine similarity among the corresponding vectors

From: Graph-based exploration and clustering analysis of semantic spaces

Only the words that are also in WordNet are considered.

Cosine similarity threshold

0.5

0.55

0.6

0.65

0.7

Number of nodes

58186

50576

39363

26509

14694

Number of edges

2033297

801085

373458

78227

19557

Average degree

69.88

31.67

13.84

5.9

2.66

Largest connected component characteristics

Number of nodes

57102

46717

29374

11363

1739

Number of edges

2032530

798363

263731

62834

5885

Average degree

71.19

34.17

17.95

11.05

6.76

Diameter

21

27

45

67

23

Average distance

5.75

7.89

11.45

18.09

8.65

Global clustering coefficient

0.43

0.43

0.41

0.39

0.28

Average local clustering coefficient

0.37

0.36

0.36

0.36

0.32

Degree assortativity

0.43

0.41

0.40

0.38

0.11

Largest clique size

245

155

89

37

14