Entity-Level Sentiment and Emotion Analytics For Decision Support In Airlines

Airlines increasingly rely on social media analytics to monitor customer feed-back, particularly during crises, when message volume increases. To help struc-ture and handle such messages, this paper proposes the Multidimensional Mod-el for Entity Recognition and Annotation (M2ERA), a three-layered framework that combines primary intent classification, contextual attributes, and entity-level sentiment and emotion annotation using a locally deployed large language model. Applied to a multilingual Facebook dataset collected during a 10-day pi-lot strike, the model enables real-time, granular analysis and integration with Social Network Analysis metrics. Results show that entity-level annotation en-hances interpretability, supports dynamic dashboards, and improves decision-making during high-pressure operational contexts.