Feedback Analysis and Management of International Travelers Using LLM Powered AI Text Analysis: A Case Study in Japanese

With the globalization of tourism and the rise of multilingual feedback, extracting actionable insights from diverse, unstructured traveler reviews has become a key challenge for public governance. This study aims to address the limitations of traditional feedback systems that rely on fragmented ratings and unstructured text, which often hinder real-time, culturally aware decision-making. We introduce a domain-adapted BERT (Bidirectional Encoder Representation from Transformers)-based AI (Artificial Intelligence) system that analyzes Japanese tourist reviews—selected for their linguistic ambiguity and cultural richness—via a multi-perspective prompt engineering framework. Our system integrates sentiment analysis, stakeholder-specific keyword extraction, and misinterpretation detection into a closed-loop feedback workflow. Compared to conventional methods, our approach offers higher accuracy, improved contextual understanding, and scalable management applications. Experimental results show that with 1,000 reviews, sentiment classification reaches 89.8 percent accuracy, stakeholder routing achieves 83.2 percent, and cultural misinterpretation detection yields 77.5 percent precision. These findings suggest that LLM (Large Language Model)-powered systems can significantly enhance tourism governance and service quality, especially when sufficient data is available. This work contributes a practical, scalable solution to AI-driven multilingual feedback analysis in public administration.