Scope is as reference (Q54828450), as qualifier (Q54828449): the property must be used by specified way only (Help)

Exceptions are possible as rare values may exist. Exceptions can be specified using exception to constraint (P2303).
List of violations of this constraint: Database reports/Constraint violations/P887#Scope, SPARQL

Value type “algorithm (Q8366), reasoning (Q1156402)”: This property should use items as value that contain property “instance of (P31), subclass of (P279)”. On these, the value for instance of (P31), subclass of (P279) should be an item that uses subclass of (P279) with value algorithm (Q8366), reasoning (Q1156402) (or a subclass thereof). (Help)

Exceptions are possible as rare values may exist. Exceptions can be specified using exception to constraint (P2303).
List of violations of this constraint: Database reports/Constraint violations/P887#Value type Q8366, Q1156402, SPARQL

Allowed entity types are Wikibase item (Q29934200), Wikibase property (Q29934218), Wikibase lexeme (Q51885771), Wikibase form (Q54285143), Wikibase sense (Q54285715), Wikibase MediaInfo (Q59712033): the property may only be used on a certain entity type (Help)

Exceptions are possible as rare values may exist. Exceptions can be specified using exception to constraint (P2303).
List of violations of this constraint: Database reports/Constraint violations/P887#Entity types

None of title (Q783521): value must not be any of the specified items.
Replacement property:
Replacement values: inferred from title (Q69652283) (Help)

Exceptions are possible as rare values may exist. Exceptions can be specified using exception to constraint (P2303).
List of violations of this constraint: Database reports/Constraint violations/P887#none of, SPARQL

None of abstract (Q333291): value must not be any of the specified items.
Replacement property:
Replacement values: inferred from abstract (Q69653744) (Help)

Exceptions are possible as rare values may exist. Exceptions can be specified using exception to constraint (P2303).
List of violations of this constraint: Database reports/Constraint violations/P887#none of, SPARQL

None of pronoun (Q36224): value must not be any of the specified items.
Replacement property:
Replacement values: inferred from pronoun used (Q73168402) (Help)

Exceptions are possible as rare values may exist. Exceptions can be specified using exception to constraint (P2303).
List of violations of this constraint: Database reports/Constraint violations/P887#none of, SPARQL

None of inference (Q408386): value must not be any of the specified items.
Replacement property:
Replacement values: (Help)

Exceptions are possible as rare values may exist. Exceptions can be specified using exception to constraint (P2303).
List of violations of this constraint: Database reports/Constraint violations/P887#none of, SPARQL

None of inferred from attribute of entity (Q105958742): value must not be any of the specified items.
Replacement property:
Replacement values: (Help)

Exceptions are possible as rare values may exist. Exceptions can be specified using exception to constraint (P2303).
List of violations of this constraint: Database reports/Constraint violations/P887#none of, SPARQL

Value record linkage (Q1266546) will be automatically replaced to value record linkage (Q1266546) and moved to determination method or standard (P459) property. Testing: TODO list

Lefima (Q113708) doesn't appear in the constraint violation report but it has type of business entity (Q1269299) as value of based on heuristic (P887), and in the item page there's warning that says "The value for based on heuristic should be personal name.".--Malore (talk) 23:48, 13 May 2018 (UTC)[reply]

The constraint reports by User:KrBot only check violations in main values, but not in qualifiers or references. They are pretty useless for source-only properties like this one. SPARQL queries can help to identify wrong use, and the official WMDE implementation of the constraints system as well. —MisterSynergy (talk) 09:27, 13 July 2018 (UTC)[reply]

I'm currently working on adding birthdays to certain (usually living) people, and sometimes I can only find a range of possible dates in the form of their astrological signs. Here's one example of how I've already used it. LadiesMakingComics (talk) 17:08, 28 March 2019 (UTC)[reply]

I'm going to add subclass-of statements with references based on heuristic (P887) based on external hierarchical database IDs, and I don't see any reason for a one-of constraint. Rather I think the heuristic itself should be made a WD item. This way heuristics can be classified too. --SCIdude (talk) 06:39, 30 September 2019 (UTC)[reply]

Yes, I agree that the one-of constraint (Q21510859) is problematic—as practically always when there is no definite amount of items to be used with a property. On the other hand, a reference property such as this one is very difficult to maintain in a usable shape if people can select any item as values; one-of constraint values are also used as value suggestions meanwhile.
Yet, I think your suggestion to define and create dedicated "heuristics items" and use a value-type constraint (Q21510865) instead of one-of constraint is a good one. Those items could also carry a somewhat instructive description about how the heuristic actually works. —MisterSynergy (talk) 06:52, 30 September 2019 (UTC)[reply]

Unfortunately I could not find an ontology of heuristics. Looking at current usage there is always something from which the claim is inferred. The algorithm may be simple or complex. Complex algorithms should be documented (github/gist) and referenced. These information bits could be collected on specific instances of heuristics so we can then put the constraint on that. --SCIdude (talk) 07:03, 30 September 2019 (UTC)[reply]

Fortunately others use heuristics on a large scale, let's learn from them:

• https://www.uniprot.org/help/unirule
• https://www.uniprot.org/help/saas

But you're right, short descriptions for heuristics may be the best start. Example: inferred from matronymic name (Q69266692) --SCIdude (talk) 07:21, 30 September 2019 (UTC)[reply]

I'd say that we directly instanciate heuristic (Q201413) in the heuristics items, instead of heuristic function (Q5748264). From my experience, there is often no formal algorithmical "heuristic function" involved, as users add these kind of references manually to indicate where they have derived a certain value from. —MisterSynergy (talk) 07:49, 30 September 2019 (UTC)[reply]

Anyway I'll be using such items from now and I hope others will, too, like @Daniel Mietchen: who adds a lot heuristically based refs. --SCIdude (talk) 08:01, 30 September 2019 (UTC)[reply]

Another example: hierarchical database ID heuristic (Q69274598) --SCIdude (talk) 08:11, 30 September 2019 (UTC)[reply]

I did change the constraints accordingly, and updated inferred from matronymic name (Q69266692) as suggested. —MisterSynergy (talk) 08:17, 30 September 2019 (UTC)[reply]

@MisterSynergy, SCIdude: I'm trying to understand which value I should use to replace Soweego bot stated in (P248) references as per previous discussion, but I'm quite confused due to mismatches between property constraint definitions and actual data:

• the value is expected to be instance of (P31) heuristic (Q201413);
• querying for those values via SPARQL returns 5 results, but none of them appear in the list of accepted values in the one-of constraint box;
• neither the instances of heuristic (Q201413), nor the accepted values seem to fit Soweego bot's use case, i.e., a statement generated by an AI/complex algorithm.

Do you think it makes sense if we add artificial intelligence (Q11660) to the list of accepted values for this property? According to [1], artificial intelligence (Q11660) looks like a superclass of heuristic (Q1981968), which is not the same item as heuristic (Q201413). Cheers, Hjfocs (talk) 10:22, 4 October 2019 (UTC)[reply]

Also, heuristic (Q201413) is instance of algorithm (Q8366), and heuristic (Q1981968) is subclass of algorithm (Q8366). It looks to me quite inconsistent, maybe not from a strict ontology modeling perspective, but at least from a terminological one. In other words, I can't see the difference between these two homonymous items. --Hjfocs (talk) 11:07, 4 October 2019 (UTC)[reply]

Hjfocs I think User:MisterSynergy just added the value constraint on top of the previous one-of constraint because removing the one-of would create lots of conflicts. But then, it should, if we agree that the value constraint is the right thing to do, no? So I'll remove the one-of now. As to heuristic, I think there are indeed two concepts (no reference sorry) but also agree that instance-of algorithm is wrong in any case. Finally, AI can be rule-based (we make WD for them) or not (like neural nets). Maybe we should just demand the value instance or subclass of algorithm. --SCIdude (talk) 12:58, 4 October 2019 (UTC)[reply]

I also created two rule items so that at least the property examples no longer conflict. --SCIdude (talk) 13:08, 4 October 2019 (UTC)[reply]

SCIdude, thanks for the quick reply and the edits. I think the property constraints make more sense now. I'll safely go ahead with artificial intelligence (Q11660) as the value for Soweego bot references. --Hjfocs (talk) 13:40, 4 October 2019 (UTC)[reply]

I've created inferred from pronoun used (Q73168402) as a value for this property, since I seem to create lots of statements with constraint violations for basing gender on pronouns used in references. - PKM (talk) 21:31, 1 November 2019 (UTC)[reply]

Thanks, I use this heuristic a lot and it's nice to be able to declare it on the statement. Ghouston (talk) 01:50, 19 November 2019 (UTC)[reply]

• I also added inferred from honorific (Q75695770) for cases like "Mrs Smith". Ghouston (talk) 09:53, 19 November 2019 (UTC)[reply]

Is it possible to write a date when the conclusion was drawn? We have retrieved (P813). Should we use that one? Or should we use another property. I think it is important to know if the value is still accurate. --Hannolans (talk) 08:45, 6 March 2020 (UTC)[reply]

I've asked for some thoughts over at Help_talk:Sources#Heuristic_references - briefly, what guidance should we have about when and how to use heuristics like inferred from image (Q105573271) (responses there ideally to keep them together). T.Shafee(evo&evo) (talk) 23:28, 5 October 2022 (UTC)[reply]