Evidence from cognitive neuroscience indicates that subcortical regions, especially the pulvinar region of the thalamus, are involved in semantic processing. In the current study, graph-based methods were used to investigate whether a cortical-subcortical network is involved in vocabulary processing. In addition to traditional resting-state functional connectivity (rsFC) analysis between local brain areas, we applied a novel method to validate the interaction between semantic network hubs and the pulvinar. Unlike the traditional rsFC, the new metrics assessed rsFC pattern similarity (rsFCS), which was calculated with a cosine similarity algorithm based on whole-network topological information. We also applied a support vector regression program based on left pulvinar connectivity patterns. A brain-behavior analysis was conducted based on 100 randomly selected unrelated participants from the Human Connectome Project S1200 database. After controlling for the visuospatial and attention test scores, the rsFC between the left middle temporal gyrus, left inferior parietal lobule, and left pulvinar was significantly positively correlated with age-adjusted picture vocabulary scores. Similar results were confirmed based on the new rsFCS analysis. The support vector regression procedures also showed a clearly relationship between picture vocabulary scores and left pulvinar-related rsFCs. Our study verified a role for a subcortical-cortical network in vocabulary processing that is based on local and whole-network topologies.