Graph-RAG Framework for Personalized Nutrition Recommendations in India

A 95.9 KB PDF authored by Varrsan Dindukurthi, last updated on 2026-04-30, describes a graph-centric decision-support framework for nutrition. The framework uses a Neo4j knowledge graph modeling diseases, nutrients, foods, and demographic-specific Recommended Dietary Allowances (RDA). It was evaluated in case studies for anemia, hypertension, and diabetes using diverse user profiles.

Use Cases

• Developing personalized meal plans based on the described graph relationships among diseases, nutrients, and foods.
• Evaluating nutrient ranking algorithms based on the deterministic cosine similarity method mentioned in the description.
• Building demographic-sensitive nutrition tools based on the framework's integration of Recommended Dietary Allowances (RDA).
• Researching explainable AI in healthcare based on the constrained language generation and graph-native reasoning approach.

Strengths

• Framework evaluated across multiple case-study scenarios including anemia, hypertension, and diabetes.
• Description details a semantic ETL pipeline for integrating heterogeneous datasets and resolving terminology inconsistencies.
• Results indicate improved ranking consistency and reduced nutrient-dominance bias compared to baseline approaches.

Limitations

• The dataset is a 95.9 KB PDF document describing a framework, not a primary data collection; actual data scale and structure are unknown.
• Column-level documentation is absent; field semantics must be inferred from the paper.
• The evaluation was conducted in a controlled setting, and the framework is described as a decision-support tool rather than a clinically validated system.

Provenance

Source

figshare, authored by Varrsan Dindukurthi.

Collection Method

Proposed framework involving a semantic Extract, Transform, Load (ETL) pipeline and a Neo4j knowledge graph.

Freshness

Last updated 2026-04-30 05:41:30; freshness should be verified.

Geography

The description mentions a culturally diverse context such as India, but specific spatial coverage of underlying data is not detailed.