A Research on Automated Customer Review Analysis through AI Hybrid Models and LangChain Framework

As a preliminary step toward quantitatively analyzing the effectiveness of online advertising media actively used in small businesses, it is necessary to automatically process and analyze customer-generated reviews. To automate customer review analysis, a combination of Instruct model(instruction-tuned) and NLI (Natinural Language Inference)–based zero-shot classifiers can be integrated within the LangChain framework to achieve optimized processing. In this study, we analyze review texts written by customers who visited small-scale medical service providers, summarize key fields, and investigate both Instruct models and NLI-based zero-shot classifiers for generating sentiment classification rationales. Based on this investigation, we construct 16 optimized model combinations. The performance of each combination is then compared and evaluated using JSON files generated through batch processing. The proposed system enables small business owners to efficiently respond to customer feedback by automatically analyzing reviews in real time. In particular, the system facilitates rapid handling of negative customer reviews, which can significantly influence the visibility and revenue of small businesses that rely heavily on online advertising media.