We present WineSensed, a large multimodal wine dataset for studying the relations between visual perception, language, and flavor. The dataset encompasses 897k images of wine labels and 824k reviews of wines curated from the Vivino platform. It has over 350k unique vintages, annotated with year, region, rating, alcohol percentage, price, and grape composition. We obtained fine-grained flavor annotations on a subset by conducting a wine-tasting experiment with 256 participants who were asked to rank wines based on their similarity in flavor, resulting in more than 5k pairwise flavor distances. We propose a low-dimensional concept embedding algorithm that combines human experience with automatic machine similarity kernels. We demonstrate that this shared concept embedding space improves upon separate embedding spaces for coarse flavor classification (alcohol percentage, country, grape, price, rating) and representing human perception of flavor.

Abstract Some grades of municipal and industrial waste glass (WG) discarded in landfills can cause environmental issues. One of the efficient solutions to overcome this issue is to use WG in concrete mixtures as aggregate or supplementary cementitious materials. Modeling the compressive strength (CS) of the produced concrete using machine learning methods can provide helpful insights into the effects of WG on concrete properties. In this study, a comprehensive database of concrete containing WG (CCWG) was gathered from 24 peer-reviewed papers. Two different scenarios were considered in the selection of input variables, and a novel machine learning method, called multi-objective multi-biogeography-based programming, was used to predict the CS of CCWG. This algorithm can automatically select the effective input variables, the structure of the equations, and its coefficients. Moreover, the proposed model optimizes the precision and complexity of the developed models simultaneously. The definition of complexity in the optimization problem can help achieve different mathematical equations with various accuracies and assist users in predicting the CS of CCWG even with a limited number of optimal input variables. The results show that the proposed algorithm can introduce several equations with different accuracies, complexities, and input variables to predict the CS of CCWG.