Quality facet - Nutrition - Flag products where 3 or more nutritional values are identical

[teolemon]

What

• Flag products where 3 or more nutritional values are identical. This is a clear sign that someone entered bogus data.

https://dk-en.openfoodfacts.org/cgi/product.pl?type=edit&code=3760265420094

Part of

• 🧽 Data quality (tracker) #5538

[benbenben2]

Sharing some thoughts about this opened issue.

3 identical values is probably not enough because it is quite easy to find products with 3 identical values in the nutrition table and that are valid.

4 seems also not enough

Even 5 identical values can be correct:

• https://world.openfoodfacts.org/product/0001200000005/aceite-coco

when almost all values are identical:

• it is often non food product (nb of identical values):
  • https://world.openfoodfacts.org/product/9787560276236/funda-de-movil
  • https://world.openfoodfacts.org/product/9788434237353/prueba
  • https://world.openfoodfacts.org/product/9788186004807/indian-cooking (6)
  • https://world.openfoodfacts.org/product/9788490494936/aaaaaa (5)
  • https://world.openfoodfacts.org/product/9788483469323 (6)
  • https://world.openfoodfacts.org/product/9788469834862/otro-mundo-a-la-vuelta-de-la-esquina-anaya (6)
  • https://world.openfoodfacts.org/product/9788467593396/cuaderno-de-expresion-plastica-sm (6)
• or it seems to be done by contributors when the product does not possess nutrition table:
  • https://world.openfoodfacts.org/product/9788175257665/test-green-fresh (6)
  • https://world.openfoodfacts.org/product/9788469628454/pan (6)
• it may be due to difficulty to extract data from the u.s./u.k. version of the nutritional fact (with 0%)

Maybe the alert could be set for 6 identical values: fat_100g, saturated-fat_100g, carbohydrates_100g, sugars_100g, proteins_100g, salt_100g.
Without fiber that is often missing and without sodium that can be either identical to salt or automatically converted from salt value.
That would include ~25 000 products

[alexgarel]

@benbenben2 what about ignoring some typical values in this comparison ? (like 0, 0.5, 0.1, 1, 0.01) which are values typically found to mean "small enough".

[benbenben2]

Yes! Good point.
That could make 2 different quality facets:

• 6 identical values (most of the time intentionally done, together with a non-food picture)
• 3 identical values + value is above 1 (to avoid 0.5 or 1 for "less than 0.5" or "less than 1"). Some examples.