ROMJIST Volume 29, No. 4, 2026, pp. 325-336, DOI: 10.59277/ROMJIST.2026.4.04
Elena-Ruxandra LUTAN, Costin BADICA Modeling Fragrance Discovery through Multi-Attribute Fusion of Olfactory Accords and Affective Profiles
ABSTRACT: This paper presents a multi-dimensional model for fragrance discovery, designed to bridge the gap between crowd-sourced olfactory and affective profiles. Traditional Information Retrieval in the olfactory domain often suffers from a vocabulary mismatch, where chemical descriptors (accords) do not align with how users experience scent emotionally. The proposed model addresses this gap by integrating fragrance accords and emotional profiles extracted from user reviews into a unified multi-dimensional representation. Multidimensional similarity is computed using domain-appropriate measures: Fuzzy Jaccard Similarity for accord matching (accommodating the inherent vagueness of olfactory perception) and Cosine Similarity for affective alignment. Unlike deterministic systems, a probabilistic re-ranking strategy is employed to ensure discovery and diversity in search results. The system is validated using a stratified k-fold cross-validation, measuring the alignment between retrieval candidates and user preferences across scent, emotion and lifestyle characteristics. The experimental results demonstrate that integrating affective perceptions with olfactory profile significantly improves the precision of intent-based fragrance retrieval.KEYWORDS: Affective computing; fuzzy representation; olfactory informaticsRead full text (pdf)