Conceptual frameworks—like 2D maps and matrices—help structure ideas, uncover patterns, and drive creative thinking. Yet defining the axes that organize these spaces is often subjective, cognitively demanding, and a barrier to broader use. We introduce LLM-Based Semantic Mapping, a method that turns arbitrary term sets into interpretable 2D semantic maps. A large language model extracts shared features and scores each term. Principal Component Analysis (PCA) then projects the results into two dimensions, with axis labels generated from component loadings. This approach automates the most effortful steps of framework creation while preserving clarity and interpretability. Users can visualize conceptual structures, discover latent relationships, and explore underrepresented idea spaces. We contribute a fully automated pipeline, a technique for axis verbalization, and a browser-based implementation suitable for ideation, analysis, and design research.