There are numerous apps that use artificial intelligence (AI) to assist wine drinkers in making their choices.
Machine learning is increasingly using multimodal data where a number of different types of data are used in training.
This typically involves text and images; sound is also common, but the use of other sensory inputs, such as the sense of taste, is a relatively new development.
Now, researchers have added human impressions of flavour as an additional parameter to their wine advice algorithm.
The researchers, from the Technical University of Denmark (DTU), the University of Copenhagen, and Caltech in the US, held wine tastings with more than 250 participants.
The volunteers were asked to arrange small cups of different wines based on how similar they thought they tasted.
The more different they tasted, the further apart they were placed – a technique commonly used in consumer taste tests.
The arrangements of the cups were photographed and then digitised in order to incorporate them into a huge data set.
Adding taste data makes more accurate predictions of users’ preferences
As well as information from the taste test, the data set included hundreds of thousands of user reviews and wine labels provided by the global wine app and marketplace Vivino.
Study leader and DTU graduate student Thoranna Bender said that the data on taste offered new options for AI and apps to provide advice to users.
It could recommend other wines that tasted similar to an already favoured bottle, for example, or combine those recommendations with price comparisons.
Co-author Professor Serge Belongie, who heads the Pioneer Centre for AI at the University of Copenhagen, added that combining the different strands of data allowed the algorithm to make more accurate predictions of people’s preferences.
He said that enabling machine learning to incorporate human sensory experiences would result in better algorithms that ultimately benefit the user.
He also said that understanding taste is a key aspect of food science and a vital part in developing food sustainability.
The use of AI in this context is currently in its infancy, but Belongie predicted that their wine tasting project would inspire more research “at the intersection of food science and AI”.
Bender pointed out that their technique could just as easily be applied to other products such as beer or coffee, or it could potentially be used to develop foods for different taste profiles.
Today’s news was brought to you by TD SYNNEX – the UK’s number one solutions distributor.
