Coursera project: NLP techniques (Word Embeddings using Word2Vec) to generate word embeddings for ingredients in recipes
These word embeddings can be used for recommendations in an online store based on added items in a basket, or to suggest alternative items as replacements when stock is limited.
- Task 1: Introduction
- Task 2: Exploratory Data Analysis and Preprocessing
- preprocess a text dataset comprising recipes
- prepare the data for use in a word embedding model
- Task 3: Word2Vec Theory and Training
- implement Word2Vec model using Gensim
- Task 4: Basic Model Analysis
- visualizing the results using a similarity matrix,
- Task 5: Building Interactive Network Graph of Results
- build a network graph using NetworkX on top of this,
- build a visual tool to explore this data in a manner that is both interactive and aesthetically unmatched, using Plotly.
This is the dataset we will be using: https://eightportions.com/datasets/Recipes/#fn:1 It is collated by Ryan Lee, sourced from Food Network https://www.foodnetwork.com/ Epicurious https://www.epicurious.com/ and Allrecipes https://www.allrecipes.com/
Use a recipe dataset, to train a model to learn the interactions between different kind of ingredients and available products in a supermarket.
- This model can then be implemented in a number of different ways, for example:
- to recommend products based on items added to cart;
- to offer alternatives products based on stock;
- to discover new products to create different recipes.