List and lexical-semantic analysis of 600 predefined keywords for rule-based NLP to extract themes within and beyond R-strategies in circular economy definitions

Thematic identification was applied in this study through a combination of deductive and abductive reasoning to ensure both structured categorization and flexibility in capturing emerging concepts. The process began with a deductive approach, selecting initial keywords for each circular economy strategy based on the 10R framework. These keywords formed the foundation for strategy classification. An iterative abductive refinement followed, where definitions were manually reviewed to identify missing or alternative terms, which were then systematically integrated. During this process, the initial keyword set was used as input for the previously developed NLP model. The model was continuously tested and refined, enabling data-driven expansion while maintaining alignment with the 10R framework.Moreover, a linguistic categorization analysis (lexical-semantic analysis (Hua et al., 2015)) was conducted, classifying all keywords into three groups: lexical matches (direct variations of the strategy name), semantic equivalents (synonyms), and contextual inferences (concepts requiring interpretation).