. Author manuscript; available in PMC: 2011 Dec 1.

Published in final edited form as: J Biomed Inform. 2010 Aug 17;43(6):1009–1019. doi: 10.1016/j.jbi.2010.08.005

Table 1.

Summary of algorithm traversing the dependency graph from entities through root to other entities. The subject and object may be modified entities, and the dependency graph provides data types that help decide how to construct the subject, object and relationship for each graph. In particular, depending on the type associated with each edge in the graph during traversal of a DG, the seed expansion either (i) continues, (ii) ends and thereby establishes a subject or an object, or (iii) interrupts. To identify a relationship, the expansion of the two seeds has to end (one as a subject and one as an object) on a common “root” word.

	ALGORITHM ACTIONS
	(i) expand seed	(ii) end expansion		(iii) interrupt
	(i) expand seed	expanded seed is subject	expanded seed is object	(iii) interrupt

*DEPENDENCY TYPES*	nn (noun modifier)	*nsubj* (nominal subject)	*dobj* (direct object) *iobj* (indirect object)	*All others*
	*prep*_{for,in,into,of,on,to} (preposition)	*nsubjpass* (passive nominal subject)	*xcomp* (open clausal component)
	*prep*_{for,in,into,of,on,to} (preposition)	*xsubj* (controlling subject)	*prep*_{at,as,by,for,in,into,on,than,with,within} (preposition)