Applying Word2Vec to Cooking Recipes
In this small notebook I have used recipes to create vector representations of recipe ingredients. Both CBOW and Skipgram Word2vec methods were used to predict ingredients within a recipe based on other ingredients. Vectors for different ingredients were clustered and visualized using T-SNE with some natural cuisine separations being visible on the graph.
Graphs in this notebook are interactive, Please use following link to get a properly rendered view: https://nbviewer.jupyter.org/github/trosenflanz/Ingredient2Vec/blob/master/Ingredient2Vec.ipynb
Data from http://www.yummly.com/ was obtained through https://www.kaggle.com/kaggle/recipe-ingredients-dataset