Exploring quantitative modelling of semantic factors for content marketing

Abstract

Developments in business analytics as well as an increased availability of data has allowed digital marketers to better understand and capitalize on consumer behavior to maximize the engagement with marketing materials. However, because most previous studies in this field have focused on consumer behavior theory, they have been largely limited in scope due to small datasets and reliance on human-labeled data. This study aims to explore the potential of using a machine-learning language model to generate vector embeddings, representing the semantics in text, to model engagement in a quantitative way. By clustering the semantic vector embeddings, the study was able to generate datasets on different topics, on which regression models were estimated to gauge the impact of the represented variables. Many of the parameters in the models were shown to be significant, implying both explanatory potential in text semantics, as well as the presented methods’ ability to model these. This expands on theories in the literature regarding how semantic factors affect consumer perception, as well as highlighting that text semantics contains information that can help inform marketing decision-making. The paper contributes a methodology that can allow academics and marketers alike to model these semantics and thus gain insights relating to how topics and language affect consumer engagement. Further investigation into similar methods might allow digital marketers to improve their understanding how different consumers perceive and engage with their marketing content.