Improving rare disease classification using imperfect knowledge graph

Table 1 Corpora statistics

	HaoDaiFu	ChinaRe
# of documents	51,374	86,663
# of classes (diseases)	805	44
Vocabulary size	59,879	41,087
Average # of words/doc	26.7	29.7
Average # of knowledge terms/doc	10.8	4.0

A “knowledge terms” is a term appearing in medical knowledge graph (see “Acquiring knowledge features from KG entities” section)

ISSN: 1472-6947