5.5. Index construction
Given a data multigraph G, we build the following three different indices: (i) an inverted list A for storing the set of data vertex for each attribute in ai ∈ A; (ii) a trie index structure S to store features of all the data vertices V and (iii) a set of trie index structures N to store the neighborhood information of each data vertex v ∈ V. For brevity of representation, we ensemble all the three index structures into I:= {A,S,N}. During the query matching procedure (the online step), we access these indexing structures to obtain the candidate solutions for a query vertex u. Formally, for a query vertex u, the candidate solutions are a set of data vertices Cu = {v|v ∈ V} obtained by accessing A, S or N, denoted as 
respectively.
5.5.1. Attribute index
The set of vertex attributes is given by A = {a0, …, an} (section 5.3), where a data vertex v ∈ V might have a subset of A assigned to it. We now build the vertex attribute index...