Introduction

For our project we decided to look at wine data in order to create a wine recommender. We are aiming to generate wine recommendations based on text reviews of wines. Ideally, we would like to create a more effective method of recommending wines that is based on tasting notes and not user preference or ratings. Many other recommenders take user ratings (sometimes on a ten point scale, or a five start scale) and are heavily reliant on user opinions of wines. Our recommender will just look at the terms used to describe the wine by cleaning the reviews and weighting words.

In order to do this we are using a dataset we found online. More information about our dataset will be available on our next blog post.

We will take the comments for each of the wines and lemmatize, tokenize, and vectorize the words within them. Then we hope to use clustering algorithms to cluster wines based on the weighted tastings notes.

This blog will follow our progress throughout the project.