Automatic identification of wines via supervised text classification

Aim of this paper is to explore the use of text classification in a sensory analysis context. The research question on which the proposal is based considered if it is possible to automatically detect a specific product from the analysis of the reviews written by a group of users, assuming that each product has peculiar distinctive characteristics. In particular, we focused on wines’ reviews written by experts which are trained to give a professional evaluation of the different products. The task described above can be accomplished by considering the text categorisation techniques developed in the framework of Text Mining. Several algorithms have been developed in literature, but we particularly focused on supervised algorithms. Following a so called k-fold cross validation procedure, a selected algorithm is trained on a labeled subset of reviews, then is tested on different subsets. In this way it is possible to estimate how accurately the predictive model performs. The results obtained by analysing 130.000 reviews scraped from WineEnthusiast magazine are presented and discussed, showing the effectiveness of our proposal.