Counterfactual Recipe Generation: Exploring Compositional Generalization in a Realistic Scenario

TL;DR

This paper explores whether pretrained language models can perform compositional generalization in recipe generation by modifying recipes based on ingredient changes, revealing current models' limitations in understanding and applying culinary knowledge.

Contribution

The study introduces a counterfactual recipe generation task, a large-scale Chinese recipe dataset, and evaluates pretrained models' ability to perform compositional modifications.

Findings

Models struggle to modify ingredients while maintaining style.

Models often miss actions related to changing ingredients.

Pretrained models do not fully learn or utilize culinary knowledge.

Abstract

People can acquire knowledge in an unsupervised manner by reading, and compose the knowledge to make novel combinations. In this paper, we investigate whether pretrained language models can perform compositional generalization in a realistic setting: recipe generation. We design the counterfactual recipe generation task, which asks models to modify a base recipe according to the change of an ingredient. This task requires compositional generalization at two levels: the surface level of incorporating the new ingredient into the base recipe, and the deeper level of adjusting actions related to the changing ingredient. We collect a large-scale recipe dataset in Chinese for models to learn culinary knowledge, and a subset of action-level fine-grained annotations for evaluation. We finetune pretrained language models on the recipe corpus, and use unsupervised counterfactual generation methods to generate modified recipes. Results show that existing models have difficulties in modifying the ingredients while preserving the original text style, and often miss actions that need to be adjusted. Although pretrained language models can generate fluent recipe texts, they fail to truly learn and use the culinary knowledge in a compositional way. Code and data are available at https://github.com/xxxiaol/counterfactual-recipe-generation.