offCanada

/

reference-db-system

Like 0

Follow

Open Food Facts in Canada 21

I like this example,
https://huggingface.co/datasets/offCanada/compliments-brand/blob/main/qa/peanut_butter_summary.parquet

I added a curated QA dataset with concrete examples for the variant-level cases:
Jelly Powder: Blueberry, Cherry, Lemon, Lime, Raspberry, and Strawberry — all 85 g
Crumb Cake: Apple, Blueberry, and Raspberry — all 465 g
Pie Filling: Apple, Blueberry, and Cherry — all 540 ml
Peanut Butter: Smooth vs. Crunchy, with 500 g, 1 kg, and 2 kg sizes
Salsa Chunky: Mild vs. Medium

Important clarification based on the current pipeline:
Package size is encoded as a variant in Phase 4. Products with different sizes can share the same group_id and have different variant_ids.
Flavour and formulation are currently part of the identity in Phase 3, so they create separate group_ids.

So, when I mentioned flavour as a variant-level difference, I meant it from the product-family perspective. In the current implementation, flavour is not stored as a variant_id.

Dataset:
https://huggingface.co/datasets/saraNour/compliments-reference-db-qa/tree/main