Re: Action: agent vs participant vs object

On 10/22/2014 12:18 AM, Sam Goto wrote:
> 
> 
> On Mon, Oct 20, 2014 at 4:48 PM, ☮ elf Pavlik ☮
> <perpetual-tripper@wwelves.org <mailto:perpetual-tripper@wwelves.org>>
> wrote:
> 
>     Hi,
> 
>     I keep reviewing examples of Action subtypes, IMO some of them use
>     agent, participant and object in possibly deceiving way.
> 
> 
> That's probably my fault :( Sorry :( I'd be happy to clarify where needed.
No worries + my apologies for using possible bit to strong term! I
shouldn't go through those examples late at night ;)


>  
> 
>     According to
>     definitions:
> 
>     * agent - The direct performer or driver of the action (animate or
>     inanimate). e.g. *John* wrote a book.
>     * particpant - Other co-agents that participated in the action
>     indirectly. e.g. John wrote a book with *Steve*.
>     * object - The object upon the action is carried out, whose state is
>     kept intact or changed. Also known as the semantic roles patient,
>     affected or undergoer (which change their state) or theme (which
>     doesn't). e.g. John read *a book*. 
> 
> 
>     I consider description of participant already including ambiguous
>     example. If John and Steve wrote book together I would see both of them
>     as agent in this action.
> 
> 
> The original attempt for "participant" were for completed actions of the
> style "X person did Y WITH Z", in that Z in "WITH Z" is the
> "participant" (think facebook posts).
Yes, that makes a lot of sense to me!

>  
> If John and Steve wrote the book, I think you'd be correct to say that
> both are agents. I think calling one an agent (primary) and the second a
> participant (secondary) is also semantically valid and correct.
Agreed! I think one could consider all the agents and in some cases
objects as participants of given action. I also noticed many useful sub
properties in: http://schema.org/participant

I don't think that in many cases we can draw a clear line between
agent/participant/(object sometimes), but I would like to avoid someone
getting impression: "only one agent allowed" or "only one object allowed"

> 
> Another way to approach this is to have an "author" property that is a
> sub-property-of agent, to specialize the role.
Somehow I would try to keep distinction between Action and publishing
online update/status/note about this Action/Activity. For example:

* John and Steve agreed with a scholar paper claiming that P = NP!.

Such activity could get originally published by some organization, where
both John and Steve participate in, and then both of them and anyone
else could re-publish it in their social streams.

> 
> So, I think "consumers" of these frames (e.g. google) should be able to
> consume all of these combinations and understand the intent of the
> "producer" in the variety of ways that this can be expressed.
Can I somehow preview all the current examples of actions in some
existing rich snippets view?


> 
> 
>     More examples follows
> 
>     * John and Steve agreed with a scholar paper claiming that P = NP!.
>     * John and Steve disagreed with a scholar paper claiming that P = NP!,
>     resulting in another scholar paper claiming that P is in fact !=   
>        NP!.
>     * John and Steve dislike an article.
>     * John and Steve like an article.
>     * John and Steve want an ipod.
>     * John and Steve reviewed an article.
>     * John travel from the US to Brazil with Steve.
>     * John planned an exercise plan with Steve.
>     * John ran 100 miles with Steve.
> 
>     All of those examples use agent: John, participant: Steve
> 
>     I think at least in some if not most of cases above *both* John and
>     Steve could act as agent!
> 
> 
> In a social stream, you'd occasionally want a main character associated
> with the stream (e.g. John ran 100 miles, where John is the main
> character. Steve was accompanying him, but that's secondary.).
Similar thoughts arise as with my comment on P = NP! example. If 3
distinct parties want to publish or re-publish this action (John, Steve,
[ a SportsTeam]). Same action in 3 different social streams may put
emphasis on different agents/participants, not sure about making a copy
of such action, minting URI for it and swapping agent/participant for
each stream...


>  
> 
>     We could at least provide some examples with
>     multiple agents and multiple objects. Otherwise one can get impression
>     that agent has cardinality equal to one. For multiple agents we can just
>     convert some of examples from list above.
> 
> 
> Yep, I agree.
> 
> Can you come up with a few examples where you have a clear distinction
> between primary agents and mere participants?
How about in examples "John and Steve ... " we'll make them both agents
while in "John ..., with Steve" we'll make Steve a participant?

>  
> 
> 
>     For multiple objects I could write an example like:
>     * John took photo of Jane, Steve and Alice
> 
> 
> You mentioned there was confusion between "objects" too. Can you give me
> more examples of where that appears?
My apologies again, I most likely overstated this one as well :(
Currently we have only one example with multiple values for object in
http://schema.org/SelectAction I'll try to come up with few more!


>  
> 
> 
>     Does it sound reasonable?
> 
> 
> It does. Thanks for pointing it out, I agree that this isn't clear at
> the moment and more examples would help.
I'll try coming up with additional ones whenever I'll think it may help
with clarify something. I also added few more checklist items, mostly
about improving examples, to https://github.com/danbri/schemaorg/pull/15
I plan to move most of them to separate PR to keep things more focused
and manageable to merge. Do you use github by any chance?

Cheers!

Received on Tuesday, 21 October 2014 23:59:49 UTC