If there’s anything I have learned at YarcData, it’s that graphs are everywhere. Some are obvious, as seen in social networks, but others take a bit of creativity to comprehend.
Case in point: As a Bay Area resident, I love heading to Napa for wine tasting. On a recent trip to celebrate a friend’s birthday, one particular winery gave us an excellent lesson on how food like cheese and chocolate enhances a varietal’s flavor. We also learned that the coffee drinkers among us preferred red over white, because they were more accustomed to bitterness. The tasting demonstrated that palettes are more nuanced than we thought and that what we enjoy drinking is heavily influenced by what we pair our drinks with.
Overwhelmed by all the combinations that could exist between various food and drink, I decided to investigate whether any researchers developed this information into a graph. I discovered that in 2011, researchers Yong-Yeol Ahn, Sebastian E. Ahnert, James P. Bagrow, and Albert-Laszlo Barabasi, published a paper about the ‘flavor network,’ which they constructed by combining a list of flavor compounds and three online recipe datasets (two from America, one from South Korea). Since then, they have added further datasets, a process that is much easier and faster when using graph analytics technology.
Scientific American discussed their findings, and created an interactive graph of the flavor network. Each node represents a different ingredient and the edges are the connections between them in North American and Eastern cuisines. The size of the node correlates to that particular ingredient’s prominence in the pool of recipes.
The results? Ahn and his colleagues found that the food pairing hypothesis, which states that ingredients of shared flavor compounds go well together, rings more true in Western cuisines. Eastern cultures tend to prefer contrasting flavors over complimentary ones. This is why American recipes frequently include milk, butter, egg, and wheat, whereas Asian recipes employ flavors that pack more of a punch, like cilantro and soy sauce.
The flavor network is more than figuring out that bleu cheese and dark chocolate make an excellent pairing. It holds a wealth of knowledge that can be applied to everything from personal diets (What’s a delicious vegan alternative to honey?) to corporate menu planning (What ingredients will make my salads most profitable?).
Moreover, the impact that can be derived by clustering entities by attributes is not restricted to the food and drink industry. By using graph analytics to discover relationships between clusters, financial institutions can target their VIP investors, sports teams can optimize their lineups, and retailers can offer more personalized recommendations to their customers.
As for me, I can’t wait to head back up to wine country and put the flavor network to the test—in the name of graph analytics, of course!