AFS, Analytical Flavor Systems, is a small company founded in 2016. It specializes in AI tools that address one of the most complex issues in beverages and fresh foods and that is especially marked in tea: the lack of understanding of the dynamics of flavors and aromas, right down to how difficult it is to describe them.
One research site states that humans can detect over a trillion aromas. But, “if I go to a farmer’s market, I can take a picture of a really lovely mushroom, but I cannot take an exact ‘flavor image’ and show it to someone and have them understand.” [Emphasis added].
Try taking pictures of, say, four lovely Japanese senchas plus a few not so goods. Then describe the flavor differences. Or, at least try. Then, maybe read some of the ultra-flowery poetical descriptions in a specialty tea catalog. Do they help you decide which one you’ll like? AFS is one of the new firms offering solutions to these challenges of flavor identification and description. It has developed its Gastrograph AI Platform. This is accessible via a smartphone app with a clear interface and graphic displays. It’s very easy to use. It draws on testers, many professionals working for clients but basically anyone who registers. They rank food and drinks across discrete categories of sensory experience.
They rate the intensity of flavors like “earthy”, “smoky”, sugary”, and even “retro-nasal” and provide preference scores. They pick specific descriptors to sharpen their detailed impressions: is this more like elderberry or tangerine, for instance? They add demographic data about themselves. Information about the ambient environment and situational factors that may affect their evaluations, such as noise, light, and temperature.
Gastograph sorts 6oo dimensions of data into 24 categories of flavor patterns. The key here is the core set of tools of AI Machine Learning: data, data and more data. Each new assessment enables the neural network software to update the patterns and weighting factors that it has built up, very much in the way that we learn from our experience.
Gastograph teases out associations that users may not even be conscious of. For example, it can deduce that the unknown flavored tea, pastry or soup you sample contains, say, the perilla herb. The taster’s response would be “What on earth is that?” (A type of shiso.) Gastrograph in effect says you don’t need to know but here’s a profile of the sample item and others like it. It translates impressions into rigorous and exact sensory experiences.
AFS’s founder, Jason Cohen, built his expertise as founder and executive director of Penn State University’s Tea Institute. This coordinated the work of twenty researchers in five fields of specialization in Chinese, Korean and Japanese tea. He had earlier developed a tool that outperformed expert tea tasters in pinpointing a tea’s point of origin. At the start of 2019, AFS had around 20 global company customers, mostly in packaged foods. In late 2018, it raised $4 million in equity funding.
Its data base contains more than 200,000 reports build up from a million customer assessments. It is unclear how many of these are for tea – beer brewing was the most popular initial customer base. Given the accelerating growth in new flavor designs for blended teas, tea/alcohol infusions and additions of tea to foods, plus AFS’s experience in tea, Gastrograph represents exactly the type of tool the major players need.
Over 80% of new products fail because of flavor misjudgments, often based on subjective data from focus groups. Agricultural ingredients vary widely in their consistency, freshness and seasonal or soil impacts that lead to fluctuations in flavor between production batches. For any tea grower, blender, wholesaler or retailer, there’s AI coming to your smartphone soon. This is just one instance. Data, data, data…