new relation QualifiedGeneration

[jdcourcol]

There is a QualifiedAssociation, a QualifiedUsage but no QualifiedGeneration.
For instance a software execution returns multiple artefacts of the same kind. I want to be able to distinguish them. (That is a simular concept to the name of an output parameter in a function).

[MFSY]

There is indeed a prov:qualifiedGeneration property in the W3C PROV-O recommendation. It is used to provide a context about how an entity was generated (generation time, ....). It's range (the type of its value) is prov:Generation for which we already have a schema written.

Now there is a difference between prov:qualifiedUsage and prov:qualifiedGeneration:

• prov:qualifiedUsage has a prov:Activity as domain and prov:Usage as range
• prov: qualifiedGeneration has a prov:Entity as domain and prov:Generation as range

So prov: qualifiedGeneration is entity centric whereas prov:qualifiedUsage is activity centric.

We have two options here:

• Add in the prov:Entity schema a property prov:qualifiedGeneration with prov:Generation as value type. Which means that to know which entity is generated from an activity and how it was generated one has to : 1) read the prov:generated activity property but also look up all generation objects that has the current activity id as prov:activity value; 2) or consistently add a generated entity in the current activity prov:generated list whenever a generation object is created.

• Become activity centric. Link to an activity all information about what entities was used and how, what entities was generated and how. This can be achieved by introducing a reverseActivityGeneration property which has an prov:Activity as domain and a prov:Generation as range and a reverseQualifiedGeneration which has a proven:Generation as domain and an entity as rang.

"reverseActivityGeneration" : {
      "@reverse" : "http://www.w3.org/ns/prov#activity",
      "@type" : "@id",
      "@container" : "@set"
    },
"reverseQualifiedGeneration" : {
      "@reverse" : "http://www.w3.org/ns/prov#qualifiedGeneration",
      "@type" : "@id"
    }

An activity payload looks like:

{
  "@type": "prov:Activity",
  "@id": "activity_URI",
 "used":[],
 "qualifiedUsage":[],
 "generated":[],
 "reverseActivityGeneration":[
    {
      "@type":"prov:Generation"
      "reverseQualifiedGeneration": "entity_URI",
      "hadRole": "role_URI",
      "atTime":"2016-09-10T06:51:00+02:00"
    }
  ]
}

Thoughts ?

[jdcourcol]

I saw that in:

So this is what I want to express:
'''
:output
a prov:Entity;
prov:wasGeneratedBy :a;
.
:metadata
a prov:Entity;
prov:wasGeneratedBy :a;
.
:a
a prov:Activity;

prov:generated :output;
prov:qualifiedGeneration [
   a prov:Generation;
   prov:qualifiedEntity         :output;
   prov:hadRole            workflow:output;
];

prov:generated :metadata;
prov:qualifiedGeneration [
   a prov:Generation;
   prov:qualifiedEntity         :metadata;
   prov:hadRole            workflow:metadata-of-output-with-log;
];

'''

I sounds that with that approach, If I have an activity, it gives me the list of "prov:qualifiedGeneration" as a list of properties.

[MFSY]

Can I know the source of the prov:qualifiedEntity property ? I'm not seeing it in the W3C PROV-O specification.

[jdcourcol]

I think they don't have qualifiedEntity anymore

[MFSY]

I think they don't have it. We use the W3C PROV-O recommendation as reference. Now I see in the example you provided that you adopted both entity and activity centric views :

Entity centric view:

:output
a prov:Entity;
prov:wasGeneratedBy :a.

Activity centric view:

:a a prov:Activity;
prov:generated :output.

I would recommend to choose one view to ease the querying. Otherwise you'll have to check both prov:wasGeneratedBy and prov:generated to know what entities was generated by a given activity.

I propose the two following options:

1. Activity centric view: from the activity towards the generated entities (how many entities are being generated ?) with the generation context.

{
  "@type": "prov:Activity",
  "@id": "activity_URI",
 "prov used":[],
 "prov qualifiedUsage":[],
 "prov generated":[],
 "qualifiedEntityGeneration":[
    {
      "@type":"prov:Generation"
      "generatedEntity": "entity_URI",
      "prov hadRole": "role_URI",
      "prov atTime":"2016-09-10T06:51:00+02:00"
    }
  ]
}

qualifiedEntityGeneration and generatedEntity respectively define as (in the @context):

"qualifiedEntityGeneration" : {
      "@reverse" : "http://www.w3.org/ns/prov#activity",
      "@type" : "@id",
      "@container" : "@set"
    },
"generatedEntity" : {
      "@reverse" : "http://www.w3.org/ns/prov#qualifiedGeneration",
      "@type" : "@id"
    }

2. Entity centric view: from an entity towards the generating activity with the generation context

{
  "@type": "prov:Entity",
  "@id": "entity_URI",
 "prov:wasGeneratedBy":"activity_URI",
 "prov:wasDerivedFrom":[],
 "prov:qualifiedGeneration":[
    {
      "@type":"prov:Generation"
      "prov:activity": "activity_URI",
      "prov:hadRole": "role_URI",
      "prov:atTime":"2016-09-10T06:51:00+02:00"
    }
  ]
}

In this second option the entity that was used an how they used are recorded in the activity activity_URI or in other usage objects.

Thoughts ?