An innovative element of the proposed approach is the use of common cliques in graphs representing documents to create a feature vector.
for - further research - common cliques in graphs - question - relevance to disaggregating text corpus into sub-sentence graph nodes?