This is an early draft shared for community feedback. If you’d like to discuss it, you can reach me at emotionalintuition@gmail.com. A practical application of the framework is available as an interactive graphical model for learning at https://emotionalintuition.com/Emotion science lacks stable, interoperable definitions for many of its central constructs (e.g., emotion, mood, affect, feeling, goals, desires, drives, etc.). This semantic instability undermines construct validity, invites jingle-jangle fallacies, and hinders cumulative synthesis across studies and methods. Advancing psychology and the philosophy of mind demands a precise semantic foundation for emotion classification. I use clear definitions, keep levels distinct, and map relations among categories; this organizes terms, prevents construct drift, and explains systematic relations (oppositions, families, and graded boundaries) among categories. The goal is not to claim final definitions but to provide a precise, revisable specification that helps researchers pre-register hypotheses, compare results across methods, and adjudicate among competing accounts with clearer failure modes. Looking forward, the same semantic backbone also scaffolds the next frontier of AI: affective-computing systems. Without exact representations of emotion—what the labels denote, how components combine, and how states transition—intelligent systems will struggle to approximate human-like understanding or agency. A componential, multi-level taxonomy provides implementable targets and bridging rules for inference and control, providing LLM with a “cognitive” architecture that evaluates and constructs emotion-relevant states.