Look for matching values in nutrition table images

[teolemon]

Parse OCR text of all nutrition images, looking for the values stored in the nutrition table (without trying to map them, just trying to see if they exist)

https://static.openfoodfacts.org/images/products/335/227/112/0121/2.json
https://static.openfoodfacts.org/images/products/335/227/112/0121/2.jpg

"description": "Déclaration nutritionnelle\n100ml\nValeur énergétique\nMatières grasses\n- dont acides gras saturés\nGlucides\ndont sucres*\n219KJ/52kcal\n\u003c0,5 g\n\u003c0,1 g\n12g\n9,7 g\n\u003c0,5g\n\u003c0,5 g\n\u003c0,005 g\nFibres alimentaires\nProtéines\nSel\n"

Create the following quality facets

# Look for fields that are missing although the nutrition table is filled

if `acides gras saturés` in OCR AND stored_sat_fat is None AND state=nutrition_facts_completed:
	write_facet(nutrition-image-contains-sat_fat-for-product-with-no-sat_fat-value)

if `Glucides` in OCR and stored_carbohydrates is None AND state=nutrition_facts_completed:
	write_facet(nutrition-image-contains-carbohydrates-for-product-with-no-carbohydrates-value)

if `dont sucres` in OCR and stored_sugar is None AND state=nutrition_facts_completed:
	write_facet(nutrition-image-contains-sugar-for-product-with-no-sugar-value)

if `Fibres alimentaires` in OCR  and stored_fiber is None AND state=nutrition_facts_completed:
	write_facet(nutrition-image-contains-fiber-for-product-with-no-fiber-value)

if `Protéines` in OCR  and stored_proteins is None AND state=nutrition_facts_completed:
	write_facet(nutrition-image-contains-proteins-for-product-with-no-proteins-value)

if `Sel` in OCR and stored_salt is None AND state=nutrition_facts_completed:
	write_facet(nutrition-image-contains-salt-for-product-with-no-salt-value)

if `Matières grasses` in OCR and stored_fat is None AND state=nutrition_facts_completed:
	write_facet(nutrition-image-contains-fat-for-product-with-no-fat-value)

if `kJ` in OCR and stored_energy_kJ is None AND state=nutrition_facts_completed:
	write_facet(nutrition-image-contains-energy_kj-for-product-with-no-energy_kj-value)

if `kcal` in OCR and stored_energy_kcal is None AND state=nutrition_facts_completed:
	write_facet(nutrition-image-contains-energy_kcal-for-product-with-no-energy_kcal-value)

# Look for images that are probably not nutrition images

if not `kcal` or `KJ` in OCR:
	write_facet(nutrition-image-does-not-contain-kcal-or-kj)

# Compare stored values with image

if stored_salt not in OCR:
	write_facet(salt-value-not-found-in-nutrition-image)
if stored_fat not in OCR:
	write_facet(fat-value-not-found-in-nutrition-image)
if stored_protein not in OCR:
	write_facet(protein-value-not-found-in-nutrition-image)
if stored_saturated_fat not in OCR:
	write_facet(saturated_fat-value-not-found-in-nutrition-image)
if stored_fiber not in OCR:
	write_facet(fiber-value-not-found-in-nutrition-image)
if stored_carbohydrates not in OCR:
	write_facet(carbohydrates-value-not-found-in-nutrition-image)
if stored_sugar not in OCR:
	write_facet(sugar-value-not-found-in-nutrition-image)
if stored_energy_kJ not in OCR:
	write_facet(energy-value-not-found-in-nutrition-image)

This should help surface:

• ingredients stored as Nutrition images
• Big mismatches and changed products
• Opportunities to fill fibers

This should empower:

• risk indicators made by a combination of those individual facets, to prioritize checks.