Research map / Graph learning / Knowledge and temporal graphs
Knowledge-Powered Recommendation for an Improved Diet Water Footprint
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Presents a knowledge-graph-powered recommendation engine (demo) that suggests lower water-footprint, more nutritious ingredient substitutes in user recipes, built from FoodKG, USDA, FoodOn, and water-footprint PDFs.
Problem
Reducing the water footprint and improving nutrition of diets by recommending ingredient substitutes in user recipes, since food is a water-intensive sector and healthy eating is hard to plan.
Approach
Combines FoodKG, USDA nutrition data, FoodOn ontology, and water-footprint data from academic PDFs into a knowledge graph; a dense neural network predicts missing nutritional and water-footprint values from ingredient embeddings and relation type, and the system ranks lower-water candidate substitutes from the same ontology tier.
Key results
- In the demonstration example, total water footprint of a recipe drops from 20,135 m³/ton to 5,215 m³/ton after substitution.
- In the same example, fat content falls from 30.32 g to 11.84 g.
- The knowledge graph has 20,778 nodes and 13 relationship types with an average response latency of about 1 s.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Substitute ingredients with recommended alternatives | Knowledge-powered recommendation engine · Example recipe | water footprint: 20,135m³/ton to 5,215m³/ton | Helped |
| Substitute ingredients with recommended alternatives | Knowledge-powered recommendation engine · Example recipe | fat content: 30.32g to 11.84g | Helped |
Limitations
- No quantitative user study or systematic evaluation; results come from a single demonstration example.
- Water footprint data sourced from limited academic PDFs; the ingredient database is acknowledged as needing expansion.
- Missing-value predictor's accuracy is not reported.
Experimental setup
Datasets: FoodKG Recipe1M USDA nutrient database FoodOn water footprint data from academic publications
Metrics: water footprint reduction in demo example fat content change in demo example average latency
Builds on: FoodKG KGTK S2ORC PDF parser Sentence Transformers
Closest papers by meaning
- Bites of Tomorrow: Personalized Recommendations for a Healthier and Greener Plate (AAAI 2025 · similarity 0.86)
- What can large language models do for sustainable food? (ICML 2025 · similarity 0.83)
- NutriBench: A Dataset for Evaluating Large Language Models in Nutrition Estimation from Meal Descriptions (ICLR 2025 · similarity 0.81)
- Recommender Ecosystems: A Mechanism Design Perspective on Holistic Modeling and Optimization (AAAI 2024 · similarity 0.79)
- Bridging the User-side Knowledge Gap in Knowledge-aware Recommendations with Large Language Models (AAAI 2025 · similarity 0.79)
- Spatial-Temporal Knowledge Distillation for Takeaway Recommendation (AAAI 2025 · similarity 0.79)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.