Wikidata:Property proposal/listed ingredient

has listed ingredient edit

Originally proposed at Wikidata:Property proposal/Generic

Done: has listed ingredient (P4543) (Talk and documentation)

Description	substance that's listed as ingredient on the packaging of the product; use P1545 to qualify the order of the listed ingredients
Data type	Item
Domain	food product (Q951964), medication (Q12140)
Example	Coca-Cola Zero (Q937783) → water (Q283) series ordinal (P1545) 1 Coca-Cola Zero (Q937783) → sulfite ammonia caramel (Q13081740) series ordinal (P1545) 2 Coca-Cola Zero (Q937783) → pantothenic acid (Q179894) series ordinal (P1545) 3
See also	has active ingredient (P3781)

Motivation

Packing ingrident lists are important information that are standardized in a specific way by regulations. The fact that there are specific rules about what's supposed to be listed and what's not supposed to be listed warrants a specialized property that goes beyond made from material (P186) and has part(s) (P527). There was previous discussion of a more general property. ChristianKl (talk) 13:23, 4 November 2017 (UTC)[reply]

Discussion

Support Thanks for this proposal, I think by narrowing the scope it allows a clearer use case and meaning. ArthurPSmith (talk) 17:16, 4 November 2017 (UTC)[reply]

Notified participants of WikiProject Food

Notified participants of WikiProject Medicine ChristianKl (talk) 01:33, 5 November 2017 (UTC)[reply]

Comment Thanks for new property proposal, but my opinion is not to create properties if they don't bring new information and in this case, it seems to me so.--Geoide (talk) 10:06, 5 November 2017 (UTC)[reply]

@Geoide: I do consider the order of ingredients on packaging new information that's currently not captured. The ingrident list of a lot of drugs is also currently not in Wikidata and having a clear property for that purpose might motivate the import of it. ChristianKl (talk) 14:56, 5 November 2017 (UTC)[reply]

In this case:

Support. Thanks--Geoide (talk) 15:04, 5 November 2017 (UTC)[reply]

Support I have had conversations with many people about how Wikidata should capture and organize this kind of information. I support the creation of such a property for food and drugs. I do anticipate some challenges, though - "listed ingredient: is going to come short for the needs of many food and drugs. For example, food and drug regulations in the United States might be different from food regulations in the European union. Some additives and ingredients which the European Union might require listing would not be listed on products in the United States. There are various ways to list ingredients ambiguously or by odd names, or even there can be strong evidence in external sources outside the product label that products contain unlisted ingredients. If there is a Wikidata property called "listed ingredients", then that will create infrastructure and precedent to capture ingredient lists and exclude external claims or explanations of the ingredients of products. I wish that this property would not create a systematic bias in favor of manufacturer claims to the exclusion of research by consumer advocacy organizations (both government and nonprofit) which may do their factchecking research on products. I have an institutional affiliation with Consumer Reports in the United States, which for decades has identified unlisted product ingredients and lobbied for government law to change to increase consumer access to product information. I agree with their arguments. The ideal Wikidata property for me would be something like "product ingredient" and which noted whether the ingredient came from the product label or another source. I would not want to favor manufacturer claims over excellent complementary information about product ingredients, because manufacturers have a bias to present their lists of ingredients in a way that seems most favorable to their selling the product. Blue Rasberry (talk) 16:59, 5 November 2017 (UTC)[reply]

As far as I understand the business model of Consumer Reports is to use the copyright over the data they gather to make money. If Consumer Reports would be open to donate data to Wikidata I would be happy to create appropriate properties to represent the donated data.

In cases where EU products and US products have different ingridient lists, I would be great to have information about both in Wikidata.

Even when producers use creative names for substances, Wikidata doesn't list the names directly but shows whatever labels Wikipedia has for items. Coca Cola says that it contains "Caramel E150d" but the relevant Wikidata item is labeled "Sulfite ammonia caramel" and thus "Sulfite ammonia caramel" gets shown. ChristianKl (talk) 18:49, 5 November 2017 (UTC)[reply]

If and when I get appropriate data sets from Consumer Reports I will share here. Obviously I am here because I advocate open data. If you ever want to have a chat about the environment which CR and your country's Stiftung Warentest are facing then email me for a phone or video chat.

It sounds like you are comfortable having nonlisted ingredients covered with this property. Why not just call it "ingredient"? Ingredients could be qualified as listed, or more commonly, not in Wikidata as listed but rather translated into whatever term Wikidata uses for the ingredient and from whatever sources are reliable. Reliable sources could be the list of ingredients, or trade journal information, a product database, or wherever else. Blue Rasberry (talk) 18:36, 10 November 2017 (UTC)[reply]

@Bluerasberry: The existing property proposal for "ingredient" faced opposition and this is an attempt to formulate a proposal that at this point in time has broad support. One issue is that you ignore the task of defining what an ingredient happens to be. ChristianKl (✉) 13:54, 12 November 2017 (UTC)[reply]

@ChristianKl: I just read the previously proposal. Wow, this gets complicated quickly. Okay, you started with Coca Cola as an example here. There will be different ingredients listed with Mexican Coke (Q6825758) and Coca-Cola (Q2813). In that case it is manageable because we have two articles. For most food products with variation over time or by country we can qualify by time and country, I suppose.

How would we note 4-methylimidazole (Q230021), which is a soda additive reputed to be dangerous but which would be listed as caramel color (Q227816)? "Caramel color" refers to the function of the chemical, and not what it actually is, as caramel color for soda would be this chemical whereas it might be a different chemical in another food. Would "reported ingredient" work in place of "listed ingredient", in case we have any reliable report to cite for what ingredient there is?

Support "listed ingredient" anyway - we need to make progress on this. Again, I speak up because I wish to avoid favoring manufacturer reported ingredient lists over whatever other ingredient reports might exist. Blue Rasberry (talk) 13:54, 13 November 2017 (UTC)[reply]

As far as I'm concerned it's always possible to create multiple items when reality is complex and there's a reason to distinguish between mutliple entities.

One great advantage of "listed ingridient" is that the lists on a package have an order. I consider it worthwhile to be able to store that order inside of Wikidata. If we would change to "reported ingredient" then the order becomes less clear.

Apart from that your link doesn't even say that 4-methylimidazole (Q230021) is an ingredient. An egg is an ingredient. Eggs contain many different substances, that you wouldn't put on an ingredient list but that are contained in the end product. There are thousands of different proteins in an egg that don't belong in the ingredient list of a pancake.

If you want to say that a product contains 4-methylimidazole (Q230021) you can always still use has part(s) (P527). We might even add a new "contains potential harmful substance" property. ChristianKl (✉) 14:24, 13 November 2017 (UTC)[reply]

@ChristianKl: Agree I think the way you originally proposed this is best. The world is full of product ingredient lists. These are easy to find and easy to verify. The order has significance. I think "listed ingredient" is a fine stand-alone property. Other reports of ingredients can be captured in other properties which are different enough in the way that anyone accesses them to merit being distinct from this property. Blue Rasberry (talk) 18:47, 13 November 2017 (UTC)[reply]

Support David (talk) 17:58, 5 November 2017 (UTC)[reply]
@ChristianKl, Geoide, Bluerasberry, ديفيد عادل وهبة خليل 2: Done as has listed ingredient (P4543). Thierry Caro (talk) 18:55, 18 November 2017 (UTC)[reply]