Graph Summarization in Annotated Data Using Probabilistic Soft Logic

Annotation graphs, made available through the Linked Data initiative and Semantic Web, have significant scientific value. However, their increasing complexity makes it difficult to fully exploit this value. Graph summaries, which group similar entities and relations for a more abstract view on the data, can help alleviate this problem, but new methods for graph summarization are needed that handle uncertainty present within and across these sources. Here, we propose the use of probabilistic soft logic (PSL) [1] as a general framework for reasoning about annotation graphs, similarities, and the possibly confounding evidence arising from these. We show preliminary results using two simple graph summarization heuristics in PSL for a plant biology domain.