Cook-Gen: Robust Generative Modeling of Cooking Actions from Recipes

Revathy Venkataramanan,Kaushik Roy,Kanak Raj,Renjith Prasad,Yuxin Zi,Vignesh Narayanan,Amit Sheth

2023-06-02

Abstract:As people become more aware of their food choices, food computation models have become increasingly popular in assisting people in maintaining healthy eating habits. For example, food recommendation systems analyze recipe instructions to assess nutritional contents and provide recipe recommendations. The recent and remarkable successes of generative AI methods, such as auto-regressive large language models, can lead to robust methods for a more comprehensive understanding of recipes for healthy food recommendations beyond surface-level nutrition content assessments. In this study, we explore the use of generative AI methods to extend current food computation models, primarily involving the analysis of nutrition and ingredients, to also incorporate cooking actions (e.g., add salt, fry the meat, boil the vegetables, etc.). Cooking actions are notoriously hard to model using statistical learning methods due to irregular data patterns - significantly varying natural language descriptions for the same action (e.g., marinate the meat vs. marinate the meat and leave overnight) and infrequently occurring patterns (e.g., add salt occurs far more frequently than marinating the meat). The prototypical approach to handling irregular data patterns is to increase the volume of data that the model ingests by orders of magnitude. Unfortunately, in the cooking domain, these problems are further compounded with larger data volumes presenting a unique challenge that is not easily handled by simply scaling up. In this work, we propose novel aggregation-based generative AI methods, Cook-Gen, that reliably generate cooking actions from recipes, despite difficulties with irregular data patterns, while also outperforming Large Language Models and other strong baselines.

Computation and Language,Artificial Intelligence,Information Retrieval

What problem does this paper attempt to address?

The problem this paper attempts to address is how to generate cooking actions (such as adding salt, frying meat, boiling vegetables, etc.) from recipes. Current food computing models mainly focus on analyzing nutritional content and ingredients but lack modeling of cooking actions, which have significant impacts on the nutritional value of food, food safety, and the potential production of harmful substances. For example, high-temperature grilling of meat can produce carcinogenic compounds. Therefore, the authors propose a new aggregation-based generative AI method—Cook-Gen, aimed at overcoming the challenges posed by irregular data patterns (such as the high variability of natural language descriptions), thereby reliably generating cooking actions from recipe texts and outperforming large language models and other strong baseline models in performance.

Cook-Gen, by using an aggregation method to handle irregularities in data patterns, can better capture the complex relationships between cooking actions, thereby improving the accuracy and robustness of the generative model. This is of great significance for building automated AI systems that can assist people in maintaining healthy eating habits.