• paper - CIKM 2022 Proceedings
• demo - http://recipemind.korea.ac.kr

We propose a computational approach for recipe ideation, a down-stream task that helps users select and gather ingredients for creating dishes. To perform this task, we developed RecipeMind, a food affinity score prediction model that quantifies the suitability of adding an ingredient to set of other ingredients. We constructed a large-scale dataset containing ingredient co-occurrence based scores to train and evaluate RecipeMind on food affinity score prediction. Deployed in recipe ideation, RecipeMind helps the user expand an initial set of ingredients by suggesting additional ingredients. Experiments and qualitative analysis show RecipeMind’s potential in fulfilling its assistive role in cuisine domain.

• Python 3.8.12
• CUDA: 11.X
• Download and extract data.tar.gz (link, 388.4MB) at directory ./data. These files are the datasets containing ingredient n-tuplets with food affinity scores and ingredient word embeddings.
• Download and extract saved.tar.gz (link, 115.3MB) at directory ./saved. These files are the model checkpoints for each random seed (1001 ~ 1005).

conda env create -f recipemind.yml
conda activate recipemind

Run the following code,

./train_script.sh {your_session_name} {random_seed_integer}

Run the following code,

./test_script.sh {your_session_name} {random_seed_integer}

The jupyter notebook RecipeMind Post Analysis.ipynb contains the source code for deploying the trained RecipeModel model in recipe ideation scenarios starting with any number of ingredients. We provided example cells that output the ideation results and attention heatmaps for interpretation purposes. The example heatmaps are the following,

• †: Equal Contributors
• *: Corresponding Authors

@inproceedings{gim2022recipemind,
  title={RecipeMind: Guiding Ingredient Choices from Food Pairing to Recipe Completion using Cascaded Set Transformer},
  author={Gim, Mogan and Choi, Donghee and Maruyama, Kana and Choi, Jihun and Kim, Hajung and Park, Donghyeon and Kang, Jaewoo},
  booktitle={Proceedings of the 31st ACM International Conference on Information \& Knowledge Management},
  pages={3092--3102},
  year={2022}
}