Table 6 Average degree of all nodes vs. average degree of words appearing in the lists of 1000, 3000, 5000 and 10000 most frequent words (from Moby Thesaurus ll and Beautiful Data, Natural Language Corpus data book (Norvig 2009)) in WordNet synonyms network, Moby Thesaurus and Word2Vec embedding of Google News and Amazon Reviews containing WordNet words (cosine similarity threshold = 0.5)
From: Graph-based exploration and clustering analysis of semantic spaces
Lexical databases | Word2Vec embeddings | |||
|---|---|---|---|---|
Number of words | WordNet | Moby thesaurus | Google news | Amazon reviews |
Average degree of most frequent words | ||||
All Words | 7.32 | 34.52 | 69.89 | 31.63 |
1000 | 22.48 | 241.33 | 10.72 | 8.16 |
3000 | 17.58 | 175.71 | 11.08 | 15.88 |
5000 | 16.02 | 157.61 | 11.50 | 17.91 |
10000 | 14.10 | 134.01 | 14.23 | 21.80 |
Ratio of average degree of most frequent words to average degree of all words | ||||
1000 | 3.06 | 6.99 | 0.15 | 0.25 |
3000 | 2.39 | 5.08 | 0.16 | 0.50 |
5000 | 2.19 | 4.57 | 0.17 | 0.57 |
10000 | 1.92 | 3.88 | 0.20 | 0.69 |