Text-Based Content Analysis on Social Media Using Topic Modelling to Support Digital Marketing

Abstract

This study aims to create Social Media Analytics (SMA) tools to help Digital Marketers or Content Creators create content topics for creating text-based Instagram content and support digital marketing strategy. Since no SMA tools can provide topic discovery for text-based Instagram content, this research aims to make an SMA tool. The data requirements to make an SMA tool include content text, content caption text, likes, comments, upload time, and content category obtained through the Instascrapper. The method used in this study is the Topic Modelling method using the Latent Dirichlet Allocation (LDA) approach to find the most dominant topic in the content. Optical Character Recognition (OCR) performs an image transformation process to extract text from text-based Instagram content images. The results of SMA tool creation are tested on three expert users, which shows that 93% of test participants could use the SMA to find topic references, and 85% can still be used by users even though they find it difficult. Since the test result shows that SMA tools still need development, for further research, SMA tools can focus on developing the user experience to increase the value of user acceptance by paying attention to the ease of the SMA tools. Also, SMA tools can focus on target users such as Data Analysts, Business Intelligence Analysts, or others within a company to support decision-making for the marketing department.