Patentometrics in middle and low-income countries is an emergent field. Here, we propose a methodological appraisal which combines text mining (lexical similarity) and network science (bipartite network) to associate the inventive structure of research and patents by matching classification standards in a middle-income country, Colombia, as a pilot study. We sourced and processed a sample of 2900 patents and 42,800 research articles. First, we matched two research and patenting classification standards using lexical similarity and modeled their connections using a bi-partite network. Based on this matching, we identified frequent correspondence between the research fields and the patent sections. Then, we estimated the correlation between research and patenting output-proxy as a national inventive and scientific capacity. There is no overall significant relationship between inventive and scientific capacities. Global and local factors, such as the growth of knowledge specialization or bypassing the path dependency of science and inventive capacities, shed light on this misalignment.
