Modelling the sensory space of varietal wines: Mining of large, unstructured text data and visualisation of style patterns
Carlo Valente, Florian Bauer, Fritz Venter, Bruce Watson & Helene Nieuwoudt
In this article two members of the Department of Information Science along with collaborators from Stellenbosch University’s Institute for Wine Biotechnology and Distell consider the value of mining publicly available sensory descriptions of wine for the extraction of meaningful domain-specific information about the sensory properties of wine. In this paper the authors introduce a novel application of formal concept lattices, in combination with traditional statistical tests, to visualise the sensory attributes of a big data set of some 7,000 Chenin blanc and Sauvignon blanc wines. These methods enabled the identification of sensory cues for specific styles. In particular, complexity was identified as an important driver of style in hereto uncharacterised Chenin blanc.
This is the first study to apply these methods for the purpose of identifying styles within varietal wines. More generally, this collaboration between information science and wine science opens up new investigations towards better understanding the complex field of sensory science.
The full article can be viewed here.