Figure 3
From: Clustering cliques for graph-based summarization of the biomedical research literature
Work flow of the summarization processing. The system includes two phases in processing. In the first phase, semantic predications extracted from MEDLINE citations with SemRep are represented in a graph. Then four processing steps (novelty filtering to eliminate uninformative predications, node centrality filtering, frequency filtering, and clique identification) successively condense the predication graph into a graphic summary. In phase two, clique clustering, an application of the UCINET, is used to partition the summary into several clusters representing summary themes. The theme of the clusters is labeled with a metapredication.