Schematic diagram of the procedure employed in this work. The top part represents the modeling steps: (1) generating the library of gravitational collapse models; (2) randomly selecting the base models depending on the global parameters; (3) evolving the thermochemistry in the collapse model; (4) including additional chemistry with post-processing; (5) obtaining the derived quantities (e.g., the abundances of some key species); and (6) producing the synthetic spectra. The middle part is the emulation: (7a) forward emulation, from parameters to spectra; (7) backward emulation, from spectra to parameters. The sensitivity using SHAP at the bottom (8) is where we perturb the neural network input features to determine the impact on the outputs.