You will first need to crawl Cookpad and extract recipes. Refer to https://github.com/mrorii/cookbot for an example Cookpad crawler.
We assume that the recipes are saved in a file where each line is a JSON-encoded item representing a single recipe. An example recipe should look like the following:
{
"id": 5678,
"name": "mokomichi's olive oil carbonara",
"description": "foobar",
"category": 11,
"categories": [177, 11, 589, 830],
"ingredients": ["olive oil", "fettucine", "egg", "bacon", "black pepper"],
"author": 1234,
"report_count": 10,
"published_date": '13/07/29',
"image_main", "http://img.cpcdn.com/recipes/157692/280/bf45e2954bed59b344a315e0ed4b5a9f.jpg",
"images_instruction": [
"http://img.cpcdn.com/steps/11985786/136/6e389fe8b3d970740069a6836f1f5a1e.jpg",
"http://img.cpcdn.com/steps/11985787/136/85c617f79cf1037a0646076cef827518.jpg",
]
}
In the steps below, we assume that the file containing these
JSON-encoded items is saved as cookpad.json.
You need to have Python 2.7, MeCab, and a corresponding MeCab python binding installed.
Install additional dependencies via pip install -r requirements.txt
Split the data into train, dev, and test sets.
The optional parameter threshold specifies the threshold
(on the number of tsukurepos) used to determine
whether a recipe should be labeled as positive or negative.
mkdir -p data
python split_data.py --threshold 5 cookpad.json data
python classify.py data --pr_curve pr.png > results.txt
python inspect_ingredients.py --n 1000 cookpad.json > ingredient_top.txt
python inspect_report_count.py --xmax 200 cookpad.json tsukurepo_counts.png
python generate_ingredient_mapping.py --threshold 10 cookpad.json ingredient_mapping.pkl
python generate_ingredient_network.py cookpad.json ingredient_mapping.pkl \
ingredient_complement_network.pkl
python extract_network_backbone.py --alpha 0.01 \
ingredient_complement_network.pkl \
ingredient_complement_backbone.gexf