Completed Nov. 2017 as a solo side project
When I'm playing board games with friends or working on a side project I often like to enjoy a glass of wine. Ever since my parents introduced me to wine in high school (we'll pretend it wasn't actually middle school), I have been fascinated by how diverse the flavor of a single species of fruit can be. I'm hoping this first dive into wine data is not my last.
After watching the documentary Somm, I started to think about how to build an Sommelier AI. The first step in that journey seemed like scraping some wine reviews and feeding that to machine learning algorithms.
Scraping the data was easy enough. I used Python's beautiful soup library and saved the data into a json data structure. I uploaded this dataset to Kaggle to share it with the world, and it has risen to become the 6th most up-voted dataset on Kaggle at the time of writing.
I've experimented with a few applications of deep learning, but life has been too busy for me to pursue this pet project further. I think that if I worked with a handful of real sommeliers, we could could create enough data to train a machine learning algorithm to recommend wines based on a person's particular palate. Some day I hope to have time to swing back around to this project because I think it could be a lot of fun. Dozens of people have reached out to me to share projects they've made with this dataset and some of them have been impressive. There is definitely potential.