Chelsea Ickes, Soo-Yeun Lee, and Keith Cadwallader

Department of Food Science and Human Nutrition, University of Illinois at Urbana-Champaign

Flavor lexicons help both manufacturers and consumers to communicate the intricacies of flavor nuances they experience within a product. Lexicon development typically requires the use of a trained sensory panel to evaluate a representative sample set of the product category and generate terms which describe the products. Some product categories, such as rum, contain a large amount of product variation making it difficult to create a lexicon in this manner. Due to the simple standard of identify for rum, only that it be made from sugarcane by-products, there is a huge amount of product variation that would require evaluating 100 or more different rums to create a complete flavor lexicon. Our aim was to create a flavor lexicon for rum through the use of web-based product reviews to minimize the cost and allow for inclusion of a greater number of rum products.
Web-based material consisting of blogs, company descriptions and review websites, were used to amass the sensory terms.  Each rum evaluation was coded for aroma, aroma-by-mouth, and taste attributes using the NVivo software. Once all evaluations were coded, word frequency analysis was conducted on coded attributes.  Terms that appeared at least 10 times throughout the entire dataset were selected for the preliminary lexicon.  A sorting exercise was performed to group the different terms into corresponding categories. Reviews for over 1,000 different rums were utilized and terms were collected from evaluations that described an array of rums, including white, gold, aged and agricole rums.
The analysis of web-based material yielded 166 terms, sorted into 22 different categories.  The most prominent terms included vanilla, oak, caramel, fruity, molasses and baking spices. This is the first study to use web-based material for the creation of a flavor lexicon. The developed wheel demonstrates that web-based material can be used for products with large variation to create a lexicon provided enough evaluations of the product exist.