Leveraging Natural Language Processing in Persuasive Marketing

The language used in marketing communication influences consumers' attitudes towards products or services and likelihood of purchasing or recommending these to others. Knowing the personality of the consumer is important in persuasion marketing and can be inferred from the abundance of consumer information available online. This paper utilizes text classification to extract consumers' personality from electronicword-of-mouth (e-WOM) and topic modelling to identify consumers' opinions. The aim is to optimizemarketing communication through personalized messages that abide to targeted consumers' personalities. The method is based on the theory of self-congruence, stipulating that consumers are inclined to purchase a brand that reflects their own personalities. Consumer reviews are obtained from TripAdvisor and their textual part is expressed as a proportion of different discussion themes identified through topic modelling. The personality of each reviewer is recognised using the textual part of their eWOM and a deep learning model trained on labelled text using the personality model of Myers-Briggs Type Indicator (MBTI). Four XGBoost (eXtreme Gradient Boosting) classifiers are trained, one for each of the four MBTI personality traits, using as predictors the topic embeddings and output the personality type of consumers. An explainable AI technique, namely, Shapley Additive Explanations (SHAP), is used to explain how the topics discussed by consumers in eWOM are related to their personality. Patterns from each XGBoost model are collated into a table showing how topics can be exploited by marketers during advertisement message design to appeal to specific consumer personalities. Preliminary results are compared against persuasion marketing and consumer behavior literature.