Overarching
Theory
People who actually like cooking don't want to sift through online blogs to find recipes. Theoretically, only one recipe is needed because every dish has a "true form".
Mission
Database of the most average, true, recipes for every possible dish.
Methods
Obtaining recipes
- Web crawler which can scour the internet and finds tons of recipes
- Can be programmed in either C or Python... I'd like to use C if time permits for greater learning however I think that doing this in Python would be quicker
- Once HTML files of recipes are found, they need to be parsed into ingredient lists and instructions using scraping
- Finding ingredients isn't too tricky
- Instructions will be slightly more difficult however shouldn't be too difficult, just need to find ordered lists
- Possible machine learning opportunity here... would like to train a model on my MacBook (what language is optimized for apple silicon?)
- Compile a large database or data structure of recipes
- Each recipe contains a list of ingredients and an instruction set
- This alone would be a major accomplishment
Averaging of recipes
- Averaging the ingredients is easy... simply have a cutoff for which ingredients need to make it to the final recipe and then average out the amounts (normalized to the batch size of course)
- Once the list of ingredients is finalized can I use machine learning to create the instruction set based on a training set of the instruction sets.
Other things to do recipes
- Create a list of 50 most used ingredients
- Using machine learning to make more recipes
Resources:
https://www.learndatasci.com/tutorials/ultimate-guide-web-scraping-w-python-requests-and-beautifulsoup/
https://www.llama.com