NLP-enabled recipe parsing iOS app

A pet project app which parsed out recipes from a given page, written over lockdown and published on the App Store in 2020. It relied on the Open Graph protocol and, where unavailable, would look to a list of supported sites with parsing instructions for each part of a recipe. The key feature of aji was its ability to tag an ingredient (e.g., 1 tbsp of syrup) and scale it to the recipe as the user adjusts portion/serving size.

It accomplishes this by injecting JS into a UIWebView, hidden from the user, which can better interpret, traverse, and parse the DOM better than any native Swift equivalent. The ingredients of the recipe are passed to an on-edge NLTagger, from MLKit, trained on the NYT ingredient phase tagger dataset. To scale, we take what quantified ingredients we have and adjust accordingly.

Otherwise, it's a standard app using messaging/pub/sub to communicate data, De/codable for serialization, designed with UIKit and pure layout constraints.

Installation

Clone the project and pod install. As part of the theming, we must make the rootContentStack, within the Tabman pod, public as opposed to internal. Xcode will identify this line when you attempt to build. Simply unlock the file and make the change.

License

GNU GPL v3