Embedding Contexts into Recipe Ingredients

Read my article for greater explanation: https://towardsdatascience.com/embedding-contexts-into-recipe-ingredients-709a95841914

Tree-based methods and ANNs have successfully been applied to predict the type of cuisine using a list of ingredients. Converting the ingredient list to a simple bag-of-words matrix, which is essentially a one-hot-encoded matrix, gives a prediction of accuracy of 78% on the Yummly recipes dataset.

Can we use Word Embeddings to improve those results?

In this work, we use Gensim's Word2Vec implementation to convert a given list of ingredients to a fixed-length vector representation. Let's see how those vectors look:

A cleaned recipe list and context vector representation gives classification accuracy of only 65%, which is lower than the baseline 78%. Let's look in detail at some cuisines and their top ingredients, shall we?

So, since the predition accuracy fell from 78% to 65%, is Word-Embedding bad?

Probably. Probably Not. There are caveats. For starters, the dataset has imbalanced classes. Possibly, a more thorough (or maybe less thorough?) cleaning of data is needed. Maybe the vectors built by Gensim need more tuning.

Anyway, getting poor results is also good research, right?