Text Mining Analysis of Q-Grader Sensory Descriptors in Specialty Coffee Under Accelerated Storage Conditions
Frank Fernandez-Rosillo, Lenin Quiñones-Huatangari, Jonathan Alberto Campos Trigoso, Eliana Milagros Cabrejos-Barrios, Segundo G. Chavez, César R. Balcázar-ZumaetaSensory evaluation is the reference method for assessing specialty coffee quality; however, the descriptive narratives generated by certified Q Arabica Graders remain an underutilized source of information. This study developed an integrated analytical framework combining conventional sensory evaluation with natural language processing (NLP) to characterize the evolution of specialty coffee quality during accelerated storage under different packaging systems. Green and roasted coffee stored in eight packaging configurations were subjected to accelerated storage at 40, 50, and 60 °C, and sensory evaluations were performed according to the Specialty Coffee Association protocol. Textual sensory descriptions were analyzed using descriptor frequency analysis, term frequency–inverse document frequency (TF–IDF) weighting, co-occurrence networks, topic modeling, and topic prevalence analysis. The results demonstrated that the evaluated packaging–product configurations (PPCs), together with storage temperature, influenced the sensory stability of specialty coffee under accelerated storage conditions. Vacuum packaging and multilayer laminated bags more effectively preserved desirable sensory attributes and higher cup scores, whereas elevated temperatures and coffee grinding accelerated quality deterioration, leading to the progressive replacement of freshness-related descriptors by undesirable storage-related sensory characteristics. The combined application of multiple text-mining approaches consistently revealed systematic semantic changes in sensory perception that complemented conventional cup scores and provided a more comprehensive characterization of quality evolution during storage. These findings demonstrate that integrating conventional sensory evaluation with natural language processing transforms expert sensory narratives into reproducible quantitative information, providing a reproducible analytical framework for the objective characterization and comparison of sensory changes during accelerated storage of specialty coffee.