Enhancing Home Furnishing App Usability: A Text Mining Approach Using BERT and Large Language Models

As businesses strive to assess the quality of mobile APP applications and understand consumer opinions, the vast volume of user feedback presents a challenge. Manually collecting and analyzing reviews is time-consuming and labor-intensive, making automated approaches essential for extracting meaningful insights. This study employs text mining and knowledge engineering techniques to analyze user reviews of home furnishing apps from three major brands: IKEA, Test Rite, and Nitori. Specifically, we focus on sentiment classification and categorical labeling tasks by leveraging transformer-based models, including pre-trained BERT and Large Language Models (LLMs). The research emphasizes fine-tuning the BERT model and designing prompt engineering strategies for LLMs, comparing their performance in extracting insights from user feedback. By efficiently processing and analyzing large-scale reviews, this study provides a systematic framework for app evaluation, offering valuable insights to inform app optimization strategies and enhance the overall user experience.