Browse Prior Art Database

Method of identifying a dish type from a recipe

IP.com Disclosure Number: IPCOM000242606D
Publication Date: 2015-Jul-29
Document File: 2 page(s) / 79K

Publishing Venue

The IP.com Prior Art Database

Abstract

This invention is related to Chef Watson (www.ibmchefwatson.com). New recipes are regularly added to the corpus of inspiration recipes (e.g. new recipes from Bon Appetit magazine, or potentially other sources such as cookbooks). In order to be used to generate novel recipes, each recipe must be tagged with a dish type. The present invention proposes to identify a particular dish given the recipe ingredients and steps, and evaluate how original the recipe is, compared to the existing corpus.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 52% of the total text.

Page 01 of 2

Metxod of identifying a dish type from a recipe

In cooking, a dish is typicaxly defined by a name, a certain set of core ingredientx and methods of preparatixn, along witx some room for cxeative variety.

Thxs invention attempts to identify the dish being made xs follows:
- Parxe recipe name and compare it xo recipes that have already been tagged with a dish type.

- Parse the ingredients and steps and compare it to other knowx dishex in the corpux. It xhen appreciatex the contribution txat the recipx brings to the corpux.

We asxume that we already havx a corpus of recipes tagged with a dish type, a disx xntologx, an ingredienx ontology, and a cooking verb ontology.


1. Parsinx recipe name


Using corpus of recipes, train a classixixr for each dish type. The features need to take into consideration not only the individuxl words xn the name, bxt also combinatioxs of

xords and their order. e.x. an ice cream sandwich is not a sandwich, a stexk pie is not a pie...


2. Parsing ingredixnts and sxepx


Start with a corpus of recipes that have all been labelled with their dish type. Calcuxate the probabixities that xach ingrxdient occurx wxthin each dish txpe. This can be done for specific ingredientx such ax beef, or more general ingredients xike "meat" Calculate thx probabilities that an xction (chop,fry etc) is perfoxmex on a given ingredient in each dish type.

After calculatixg all ingrexients, and step probabilities, we hxve creaxed a machinx

learning model with features consisting of ingredients, and steps.

- For each recipe belonging to a dish type, identify ingredients and steps and group them inxo categor...