Extract origins of ingredients

[teolemon]

What

• Some geographical information is not in the ingredient list, but spread on the packaging
• Extract this information to detections, that can then be combined to create insights.
• This will help get origins for the eco-score computation.
• They can be asked as the following Yes/No question: "Is the following statement valid ?"

Triggers

• “Importé du {Pérou} >> en:Ingredients from Peru
• “{Quinoa} origine non-UE” (séparé de la liste des ingrédients) >> en:Quinoa from outside Europe
• “Certains ingrédients ne proviennent pas de {France}” >>en:Non {{French}} ingredients
• “à partir de blé dur de diverses origines” >> en:Wheat from various origins

Part of

• OCR/Regex based detections #600
• Support the Eco-Score through increased data extraction (tracker) #376

[senthilnathan-ra]

@teolemon I'm new to the community and want to contribute. Is this issue still open?

[Jagrutiti]

Hey @teolemon

How do I get the geographical information? Is there an API for it? Could you please shed more light on this?

Thanks :)

[alexgarel]

@Jagrutiti informations are to be searched in OCR.
There already are such kind of regexp based search in OCR. For example for nutrients : https://github.com/openfoodfacts/robotoff/blob/master/robotoff/prediction/ocr/nutrient.py (and the folder contains more).

It's something like building regular expressions (in diverse language) expecting a country in it (we then have a taxonomy for countries, so you can get all known countries from there : https://static.openfoodfacts.net/data/taxonomies/countries.full.json, such file can be uploaded by robotoff through a scheduler method)