The Computational Wine Wheel 2.0 and the TriMax Triclustering in Wineinformatics

摘要

Even with the current state of technology, data growth is increasing so fast that without proper storage and analytical techniques, it is challenging to process and analyze large datasets. This applies to knowledge bases from all fields and all kinds of data. In Wineinformatics, various kind of data related to wine, including physicochemical laboratory data and wine reviews, are analyzed by data science related researches. In the previous work, we proposed the Computational Wine Wheel, derived from 2011's top 100 wine, to automatically process and extract key attributes from human-language-format wine expert reviews. In this work, past 10 year's top 100 wines are collected and formed a 1000 excellent wines dataset to further improve the Computational Wine Wheel. The extraction process led to the creation of what we call a Computational Wine Wheel 2.0, which is a wine attribute dictionary consisting of 985 categorized and normalized wine attributes. After the Computational Wine Wheel 2.0 is formed, we experiment it on a region- and grape type- specific dataset to seek new types of information in Wineinformatics. A novel TriMax Triclustering algorithm specifically used for the dataset processed by the Computational Wine Wheel is proposed and applied to discover three dimensional clusters (Wine × Attributes × Vintage) in wine. We found that the TriMax Triclustering algorithm produced promising and cohesive results that can be used in various aspects of the wine industry, such as defined palate grouping and wine searching.