This study compared the discourse structure of ‘Plant-based milk’ between A (2015~2017) and B (2022~2024) using Korean web texts. After applying uniform preprocessing to sources from Naver, Daum, and Google, the frequency, centralities, and the structural properties of the network on the top-30 nodes were investigated. The QAP correlation was calculated on a weighted co-occurrence adjacency matrix constructed from the 17-node intersection and assessed structural reconfiguration via CONCOR. The hub formed by milk, plant-based, and protein persisted while the edges and density increased from 240 to 266 and from 0.276 to 0.306, respectively. QAP yielded r=0.793 and p=0.001, indicating significant similarity in network-wide tie-strength patterns. In addition, CONCOR showed the foregrounding of segmented product lines and market, use, and menu contexts implying a reallocation of the periphery.