How consumer opinions are affected by marketers: an empirical examination by deep learning approach

• Published in: Journal of research in interactive marketing 2022-12, Vol.16 (4), p.601-614
• Main Author: Yu, Billy
• Format: Article
• Language: English
• Subjects: Bias; Big Data; Consumers; Decision making; Deep learning; Empirical analysis; Feature extraction; Framing; Marketing; Natural language processing; Neural networks; Questionnaires; Tourism
• ISSN: 2040-7122; 2040-7130
• DOI: 10.1108/JRIM-04-2021-0106

PurposeThe natural language processing (NLP) technique enables machines to understand human language. This paper seeks to harness its power to recognise the interaction between marketers and consumers. Hence, this study aims to enhance the conceptual and future development of deep learning in interactive marketing.Design/methodology/approachThis study measures cognitive responses by using actual user postings. Following a typical NLP analysis pipeline with tailored neural network (NN) models, it presents a stylised quantitative method to manifest the underlying relation.FindingsBased on consumer-generated content (CGC) and marketer-generated content (MGC) in the tourism industry, the results reveal that marketers and consumers interact in a subtle way. This study explores beyond simple positive and negative framing, and reveals that they do not resemble each other, not even in abstract form: CGC may complement MGC, but they are incongruent. It validates and supplements preceding findings in the framing effect literature and underpins some marketing wisdom in practice.Research limitations/implicationsThis research inherits a fundamental limitation of NN model that result interpretability is low. Also, the study may capture the partial phenomenon exhibited by active reviewers; lurker-consumers may behave differently.Originality/valueThis research is among the first to explore the interactive aspect of the framing effect with state-of-the-art deep learning language model. It reveals research opportunities by using NLP-extracted latent features to assess textual opinions. It also demonstrates the accessibility of deep learning tools. Practitioners could use the described blueprint to foster their marketing initiatives.