This document describes a text-mining approach for retrieving protein networks from PubMed abstracts. Key steps include performing a PubMed query to retrieve a set of abstracts, using named entity recognition to identify proteins mentioned in the abstracts, counting protein mentions to calculate an enrichment score for each protein compared to background frequencies in the entire PubMed database, and using enrichment scores to build a set of proteins for visualization and analysis in the STRING and Cytoscape apps. The approach is described as fast and flexible but lacking provenance information.