Network Text Analysis for Marketing Intelligence: Deconstructing the 'Gamis' Word Association Network using Modularity and Centrality Metrics

Abstract

The proliferation of unstructured textual data on social media offers a rich repository for marketing intelligence, yet the extraction of actionable insights remains challenged by the high volume of noise and the limitations of frequency-based text mining. This study proposes a Network Text Analysis (NTA) framework to examine the semantic topology of online conversations regarding "Gamis" (Islamic dress) during the pre-Ramadhan peak season. Utilizing a dataset of 1,000 microblogging interactions collected between February 1–15, 2026, we constructed a weighted undirected graph to model word co-occurrence patterns. By applying Degree Centrality and the Louvain Modularity algorithm, we identified the latent structure of the discourse. Contrasting with traditional sentiment analysis, our graph-theoretical findings reveal a network heavily dominated by "transactional imperatives" (e.g., check, get, shopee) rather than organic consumer opinion, indicating a supply-side saturation of the digital space. The modularity analysis successfully partitioned the network into three distinct semantic communities: promotional buzz, product aesthetics (luxury, elegant), and situational context (Eid, party). These results demonstrate that NTA provides a superior methodological advantage over "bag-of-words" models by preserving the relational context of terms, allowing marketers to visualize the structural gaps between seller push-marketing and actual consumer preferences