Skip to main content

Table 3 Entity label-level precision and recall estimates

From: Developing automated methods for disease subtyping in UK Biobank: an exemplar study on stroke

Concept mentions

Precision (%)

Recall (%)

Metastatic tumor

93

87

Aneurysm

97

100

Intracerebral haemorrhage

95

100

Time old (temporal words/phrases indicating old events, e.g., old ischemic stroke)

78

83

Subdural haematoma

97

83

Contusion

100

100

Subarachnoid haemorrhage

90

100

Related to (words/phrases indicating relations between two events, e.g., bleeding because of a recent fall)

78

100

Ischaemic stroke

90

99

Time recent (temporal words/phrases indicating recent events, e.g., accute ischemic stroke)

91

97

Meningioma

100

100

Transformation

88

100

Traumatic

100

75

  1. Numbers are mean values of tenfold cross validation