It is a method for systematically comparing patent papers using feature graphs. The main goal is to find shared and different characteristics in these texts and represent them as nodes and edges in an undirected network. The process includes two important tasks: calculating feature linkage scores and creating the feature graph. It enables increased analysis of patent papers for applications such as intellectual property management and innovation evaluation by using natural language processing techniques and mathematical modelling. The computation of the linkage score is based on the co-occurrence of characteristics within the same phrase as well as their relative frequencies in each document. A specialized formula measures the relationship between two vertices, with a threshold deciding whether or not they are connected. Feature extraction entails pre-processing papers in order to find relevant words, whereas graph creation incorporates characteristics from both documents, allocating edges to couples with linkage scores greater than a certain threshold. This approach provides a visual way to investigate feature correlations, assisting in the detection of document similarities and differences.