Re: PROV-ISSUE-381 (jzhao): Feedback and refactoring suggestion to prov-o section 3.2 [PROV-O HTML]

Hi Jun,
thanks for your feedback. You will find my answers below:

2012/6/18 Jun Zhao <jun.zhao@zoo.ox.ac.uk>

> Hi Dani,
>
> Sorry for my late reply. I have been trying to catch up...
>
>
> Some further feedback re. section 3.2.
>
> At the end of 2nd paragraph, it says that "Further, Agents may be
> Entities, in cases where one wishes to describe the provenance of Agents."
> Based on this, I infer that an Agent can be a subclass of an Entity, but
> this is not defined as such in the current ontology. Did I misunderstand
> you?
>
Agents MAY be entities, but they may be not. With the current modeling
agents are not subclasses of entities (since not all agents are entities)
but they are not disjoint, which means that an agent can be an entity. If
we make agents subclass of entity, then we force them to be entities all
the time. (I think we discussed this in the prov-o telecons, but you may
not have been there at that time). Does this make sense now?

>
> [...]
>
>
>
>>>  I have rewritten this part, according to your suggestions. I changed the
>> categories a bit, but more or less it's like yours.
>>
>
> Thanks! I only found prov:generated was missed in both the figure and text.

prov:generated appears in the text, but not in the figure. I'll notify Tim
about this (he is the one who made the figure).

>
>
>
>>
>>> === Refactoring examples ===
>>> Some examples are a bit long, and are not going directly to demonstrate
>>> how the expanded terms can be used. And we should also be careful that we
>>> are writing a spec of the ontology, not a how-to guide.
>>>
>>
>>
>>  I am referring to the examples based on the order of their appearance in
>>> the spec.
>>>
>>> ==== Example 1 ====
>>> - Can we remove some of the setting-up-the-scene provenance statements?
>>>
>>>  I have separated the example to simplify, but I think that most of the
>> statements are either illustrating some expanded terms or necessary to
>> understand the experiment. Could you specify which ones would you like to
>> remove?
>>
>
> I meant specially those triples for defining the agents, like about Derek,
> who has been defined in the starting-point example, hasn't he? I am not
> sure people would generally have trouble to understand derek, monica,
> chartgen without all the additional triples. But it's just my personal
> opinion:)

Ah I liked those :( . I thought of every example so it could be seen
independently from the others, and I don't think they overload the example
too much, don't you think?

>
>
>>
>>> ==== Example 2 ====
>>> 1) The theme of the example seems to show off the different types of
>>> prov:tracedTo. But it did not show the difference between prov:tracedTo
>>> and
>>> prov:wasDerivedFrom. By some annotations in the example or explicit
>>> statement?
>>>
>>>  Thanks for spotting this, I have added a couple of sentences so users
>> can
>> see the difference
>>
>
> Sorry, but I couldn't find them? Above or below the updated example? I
> mean particularly about prov:tracedTo.

I have added them as comments in the example:

prov:tracedTo         :aggregatedByRegions;                      ## If
the file hadn't existed, Monica would have not written the post.

## However the file is not what Monica used as primary source for
writting the new post,

## so we can't consider the post to have "been derived from" it.

Do you think we should add extra explanations above or below, or that is
enough?

>
>
>>
>>> 2) I am not sure about the example of prov:hadOriginalSource. The current
>>> example does not show me how it is hugely different from what
>>> prov:wasDerivedFrom. DM says that prov:hadOriginalSource is meant to
>>> bring
>>> some sense of attribution to the source entity, such as a new paper is
>>> based on existing data shared by other scientists. Can we revise the
>>> example or make it clearer?
>>>
>>>  Done
>>
>
> Mmmm, have you been kept up with the discussions about issue 395. I
> understand that this original source has more of a meaning as a primary
> source, rather than a simple derivation relationship. And I am not sure
> this is very clear from the example atm.
>
> What I am not sure is how much details we need to go in the spec. It's not
> easy to make *every* term totally understandable with just one line or just
> one RDF triple. I wonder whose responsibility it is? Should the
> cross-reference section be the more appropriate place?

I have discussed this with you offline and with the extra comments plus the
rename of the property everything is now more clear.

>
>
>>
>>> 3) We need prefix to prov-o properties and concepts in this example.
>>>
>>>  I have reworked the example including the ones that were missing. Do you
>> still have this issue?
>>
>
> No more:)
>
>
>>
>>> ==== Example 3 ====
>>> Is this about Notes or Accounts? IMO, the example needs to be enriched or
>>> removed. Should we also say what RDF syntax we should to express
>>> Accounts?
>>>
>>>  It was about both, but it no longer makes sense. Thus I have replaced it
>> with some invalidation examples.
>>
>
> Great!
>
>
>>
>>> ==== Example TBD ====
>>>
>>> We don't have expanded explanation of prov:wasStartedBy and
>>> prov:wasEndedBy. In the recent discussions we revealed there there were a
>>> lot of different "trigger" scenarios:
>>> - started by a person / agent;
>>> - started by an entity;
>>> - started by an activity.
>>>
>>> Either an example or additional text is needed.
>>>
>>>  Hmm, I haven't dealt with this yet. I added an example of an activity
>> being
>> started by an agent, but I think that the complete example would fit in
>> better in the cross reference section.
>>
>
> Agree.
>
>
>>
>>> One more sentence to say that prov:generated is an inverse of
>>> prov:wasGeneratedBy? Copy the nice sentence in the ontology annotation:
>>> This inverse of prov:wasGeneratedBy is defined so that Activities being
>>> described can reference their generated outputs directly without needing
>>> to
>>> 'stop' and start describing the Entity. This helps 'Activity-centric'
>>> modeling as opposed to 'Entity-centric' modeling.?
>>>
>>
> Thanks!
>
> Again, sorry for the late reply....
>
No problem!
Cheers,
Dani

>
> -- Jun
>
>>
>>>  Done
>>
>> Best,
>> Dnaiel
>>
>>
>
>