Skip to main content

Table 4 Information about general semantic relation evaluation dataset

From: Evaluating semantic relations in neural word embeddings with biomedical and general domain knowledge bases

Subtasks # of evaluation Percentage1 Relation Average2
  terms   terms #  
UMLS synonym 9,235 41.47% 9,211 3.14
Synonym 15,591 70.01% 48,030 4.25
Antonym 2,225 9.99% 2,977 1.38
Hypernym 16,400 73.64% 58,154 4.85
Hoponym 9,168 41.17% 112,215 19.84
Holonym 4,694 21.08% 9,944 3.69
Meronym 3,191 14.33% 13,056 7.14
Sibling 13,993 62.83% 869,814 86.38
Derivation 9,620 43.20% 27,782 2.92
Pertainym 926 4.16% 987 1.20
Total 22,271 38.19%* 1,168,921 9.35
  1. 1Percentage of evaluation terms that has at least one term with the relation
  2. 2Average number of relation terms that this type of evaluation terms has
  3. *Average of all the subtasks
Back to article page