From raw food data to real intelligence.
Barliva is a five-stage pipeline. Each stage adds structure and meaning, turning scattered public records and manufacturer feeds into a canonical, multilingual food knowledge graph with explainable health intelligence.
Ingest from many sources
We pull from OpenFoodFacts, manufacturer data, product metadata and other structured feeds. No single source is treated as ground truth — each is evidence to be reconciled. See data sources →
AI normalization
Free-text fields are cleaned, parsed and translated. Ingredients are classified, additives mapped to E-numbers, allergens detected, and conflicting values reconciled with confidence scoring and audit trails.
Canonical knowledge graph
Records resolve into canonical entities — one product, one ingredient, one allergen — connected by relations and expressed across languages. Inside the engine →
Health intelligence
A category-aware baseline plus transparent penalties and capped bonuses produces an explainable 0–100 health score, with additive and allergen context. How scoring works →
Consumer insight
Finally, Barliva generates localized, plain-language explanations so anyone can understand what a product contains and why it scored the way it did.
Watch the pipeline in action.
Scan a product in the app and see every stage's output.