Automatic classification of diseases from free-text death certificates for real-time surveillance

Table 1 Types of features — both term and concept-based — extracted from death certificates

	Feature type	Description	Example certificate extract	Resulting feature values
Term	TokenStem	A token stem, i.e., the stemmedversion of a word.	Acute chronic renal failure	Acut, chronic, renal, failur.
	TokenStem n-gram	The n-gram formed by n adjacenttoken stems.	chronic renal failure	Chronic renal, renal failur.
Concept	SCTConceptId	SNOMED CT concept identifier (as extracted by the Medtex system)	chronic renal failure	90688005.

(Stemming is a process of removing and replacing word suffixes to arrive at a common root form of the word.)

ISSN: 1472-6947