AI Ads: Practicability of Text Generation for F&B Marketing

Abstract

Despite numerous studies and implementations on text generation in the field of Artificial Intelligence (AI), the purpose of this research is to study the challenges and practicality of using Natural Language Generation (NLG) on Food and Beverages (F&B) marketing text generation using generative AI techniques. Since there are a number of different researchers that claim they are able to generate realistic text that is acceptable, we would like to investigate the degree to which AI models can effectively generate marketing text for advertising purpose. The goal of this research is to investigate several existing techniques for text generation where implementation of Long Short-Term Memory (LSTM), Open Pretrained Transformer (OPT) and Keyword to Text Generation (K2T) will be experimented. Locally collected marketing text samples will be applied to train and validate based on how realistic the generated texts are from the perspective of human expert based on realistic and practical attributes. The text-based advertisement will be developed and reviewed in the current research. A discussion based on the models’ theoretical and algorithmic fundamentals has been presented from the perspective of models’ performance.