Note: The Nutriapp mobile client was built with the intention of being an Android app and has been hosted as a web version (Nutriapp Web) for easy accessibility for demonstration purposes. It is best experienced when accessed from a mobile device.
Finding recipes using keywords and filters can be tedious, and consume more time than it's worth. Natural language querying allows users to type in a sentence that covers all of their requirements, and returns a list of very real recipes. Recipes are not generated using a NL model since the information given back to the user could be invalid or false. NL queries will be used as input to a NL model that will generate a JSON object that contains the necessary fields to be used in a query that can be sent to an API that returns recipes. This allows for platform-independence and accuracy when constructing queries to the external API.
- GPT 3.5: We make use of Open AI's GPT 3.5 to parse natural language into JSON objects that can be used to construct queries to a recipe API.