Assume products don't contain palm oil if we didn't detect palm oil in ingredients, even if some ingredients are unidentified

[stephanegigandet]

Is your feature request related to a problem? Please describe.

For most products, we mark the ingredients analysis for palm oil, vegan and vegetarian as "unknown" because even though we didn't detect palm oil, meat etc. there is at least one ingredient that we didn't recognize, so there is a possibility that it contains palm oil or meat etc.

e.g. for US products: https://us.openfoodfacts.org/ingredients-analysis

Ingredients analysis Products *
Vegetarian status unknown 212254
Palm oil content unknown 176979
Non-vegan 146183
Vegan status unknown 125784
Palm oil free 84100
Vegetarian 51024
Non-vegetarian 45105
Vegan 44871
Palm oil 37854
May contain palm oil 30830
Maybe vegetarian 21732
Maybe vegan 13271

For palm oil specifically, I think we could relax that, and make the assumption that if there is palm oil listed, then we will recognize it.

It's probably ok in almost all cases (except OCR / spelling mistakes) as there are relatively few wordings for palm oil

We can't do it for vegetarian / vegan as there are countless unidentified ingredients that do contain meat.

Describe the solution you'd like

Mark products as "palm oil free" if:

• we have a list of ingredients
• we have not detected palm oil in any of the recognized ingredients
• even if there are unrecognized ingredients

A variant could be to have a threshold on unrecognized ingredients (e.g. max 20% of unrecognized ingredients). That would remove products that have bogus lists of ingredients, or lots of OCR mistakes, that would increase a lot the chance of missing a mispelled "pa1m oil"

Describe alternatives you've considered

Make ingredients analysis perfect. :)

Additional context

Discussion in Slack: https://openfoodfacts.slack.com/archives/C06A7LENM/p1650986280363959

Number of products impacted

No response

Time per product

No response

[alexgarel]

@stephanegigandet

Mark products as "palm oil free" if:

Could it be that we can add a rule "we have no generic fat ingredient" (eg ingredient list contains "huiles végétales").

[stephanegigandet]

Could it be that we can add a rule "we have no generic fat ingredient" (eg ingredient list contains "huiles végétales").

@alexgarel : If we have a generic fat ingredient, we will mark it as "may contain palm oil" (which is different thant "palm oil content unknown").

e.g. https://fr.openfoodfacts.org/cgi/test_ingredients_analysis.pl?ingredients_text=huile+v%C3%A9g%C3%A9tale&type=add&action=process&submit=Envoyer