pkuaaron / pying

Python module to parse ingredient names. Splitting them into the ingredient, unit and quantity. It is trained on a publicly available dataset using Tensorflow.

PyIng - Ingredient parser

This is a python package for parsing ingredient strings. There is only one function parse_ingredients. It takes in a single string or list of strings and returns a dictionary containing the name, quantity and unit in that recipe string.

Get Started

First install the package using pip:

pip install pying

Then you can use it like follows:

from PyIng import parse_ingredients

ingredients = ["3 large melons", "5 1/2 cups water", "2 cups flour"]

parsed_ingredients = parse_ingredients(ingredients)

the output parsed_ingredients should look like so:

parsed_ingredients = [
    {
        "name": "melons",
        "unit": None,
        "qty": 3.0
    },
    {
        "name": "water",
        "unit": "cups",
        "qty": 5.5
    },
    {
        "name": "flour",
        "unit": "cups",
        "qty": 2
    }
]

Training the model

The ingredient parser uses a LSTM model written in tensorflow. It is trained on a publically available dataset produced by the new york times:

https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=&cad=rja&uact=8&ved=2ahUKEwjowpyb9uj2AhWNiFwKHcFQBC8QFnoECAgQAQ&url=https%3A%2F%2Fgithub.com%2Fnytimes%2Fingredient-phrase-tagger&usg=AOvVaw1AHIgZ0BfSe8ddG7E8alYt

The entire model is trained using the train_model.ipynb and the model is saved in tflite format. Please try and improve on my model, it is far from optimal, I wanted to get something that works without spending too long.

THere are more details on training the model in the jupyter notebook.