Table 4.

Corpus statistics

Name	Description	Filter	#	%
	all TTE reports		70441	100.0
	only relevant sites	f site	68915	97.8
T d	dominant layouts	fsite, fchar≥800, ${\overline{f}}_{\text{li}}$	63489	90.1
T u	mostly unstructured	fsite, fchar≥100, fchar<800, ${\overline{f}}_{\text{li}}$	2712	3.9
T c	uncommon layout	fsite, fli	1041	1.5
	mostly defective	fsite, fchar<100	1673	2.4

f_site: filter that excludes three sites of the hospital. f_char≥n: require at least n non white space characters. f_li: at least 5 list elements