Development of a Sensory Lexicon and Predictive ANN Modeling for Black Queen Wine: A Novel Workflow Incorporating Bridge-Linked QDA and Consumer Hedonic Analysis

Vitis vinifera L. × Vitis labrusca L. cv. Black Queen (BQ) is a hybrid cultivar with oenological potential in subtropical climates, yet its sensory structure remains insufficiently systematized. This study aimed to construct an integrated sensory framework by merging two Balanced Complete Block Design (BCBD) datasets into a unified database and developing a structured descriptor reduction workflow to address multicollinearity and redundancy. The resulting “BQ Lexicon v.0” comprised nine Quantitative Descriptive Analysis (QDA) attributes and twelve check-all-that-apply (CATA) descriptors. Based on this optimized dataset, an Artificial Neural Network (ANN) model was developed to predict overall liking (OL), achieving a satisfactory performance (R2(train) = 0.70 and R2(validation) = 0.74). Three-dimensional response surface visualization further illustrated non-linear relationships as a process monitor, indicating sourness as a primary negative driver of acceptance and revealing interactive and synergistic effects between tannin, sweetness, and aroma. These findings demonstrate that integrating structured data management with machine learning can enhance sensory modeling efficiency. Ultimately, the validated BQ Lexicon v.0 and the aligned data framework establish a reliable foundation for future oenological research in Black Queen grape. This structured approach effectively resolves the challenges of integrating distributed sensory datasets, while offering practical insights for targeted winemaking strategies.