Semantic and Topological Patent Graphs: Analysis of Retrieval and Community Structure

Classification systems are a common way to organize knowledge. The Cooperative Patent Classification (CPC) is a hierarchical patent classification system which intends to uniformly assign at least a single class to every document. The classification system is often successfully used as a filtering mechanism to improve patent retrieval performance. Semantic information on the other hand frequently fails to do this or only helps marginally. In this work, we build two graphs to address this: a semantic graph out of the full textual content of the patents and a topological graph out of the classification system. The semantic graph is then compared against the topological graph. This provides a basis on when semantic retrieval techniques can be useful in patent retrieval. In this work, we further visualize search result graphs and their communities to represent classification information before search result filtering.