All of human cooking compressed into 2 megabytes (arxiv.org)

Article summary

Researchers have developed a system called Epicure, which uses ingredient embeddings to represent recipes in a compact and structured way. The system is trained on a large corpus of recipes in multiple languages and can generate graphs that show the relationships between ingredients and cooking steps. This approach can be used to compress recipes into a small amount of data, making it possible to store and transmit them efficiently. The researchers have released their data and code, allowing others to build on their work.

Main themes

• Recipe representation
• Ingredient embeddings
• Cooking graphs
• Recipe compression
• Multilingual recipes
• Food science

What commenters say

• The use of graphs to represent recipes can make them easier to read and understand, especially for complex dishes.
• Some commenters have developed similar systems for representing recipes, using techniques such as dependency graphs and flowcharts.
• The graphical representation of recipes can be improved by minimizing edge crossings and using clear and consistent notation.
• The system could be useful for applications such as recipe generation, meal planning, and cooking instruction, but may not be suitable for all types of recipes or cooking styles.
• The use of a standardized format for representing recipes could facilitate the sharing and exchange of recipes across different platforms and communities.
• Some commenters have suggested that the system could be extended to include additional information, such as cooking techniques, ingredient quantities, and nutritional information.
• The graphical representation of recipes can help to reveal the underlying structure and patterns of cooking, and could be used to develop new recipes and cooking techniques.
• The system may not be suitable for all users, particularly those who prefer a more traditional or intuitive approach to cooking and recipe representation.