Tag
The paper introduces an LLM-based topic modeling method and evaluation framework that simultaneously achieves interpretability, topic specificity, and polarity stance consistency, demonstrating superior explanatory power for external outcomes like employee morale using large-scale Japanese corporate review data.