This paper introduces a comprehensive methodology for the extraction of representative information from patent papers through the application of Steiner trees. The core element of our approach is the "Extracting Representative Tree" module, designed to generate Steiner trees from feature graphs, thereby creating feature trees. Leveraging discriminative features acquired from a preceding module and the feature graph, our method constructs Steiner trees, addressing the challenge of forming coherent structures from disparate discriminative characteristics. The concept of Feature Tree Extraction is explored in depth, highlighting its pivotal role in closing gaps between discriminative features. We emphasize its utility in linking these features efficiently, thus facilitating a holistic framework for the comparative analysis of patent papers. This innovative approach promises to enhance feature-based comparative analysis within the domain of patent papers.

Steiner trees, feature extraction, discriminative features, patent papers, feature graphs, comparative analysis