APML code

28 views
Skip to first unread message

Alberto Bermejo

unread,
Nov 17, 2009, 5:29:48 AM11/17/09
to APML.Public.General
Hello, I am doing a project and I'm interested in using APML, and I
found several problems such as not being sure if you follow the
development of the standard and the other problem is that I not find
the source code, anywhere, is open source?
The only thing I have found was documentation in java and programs in
*.cs. but nothing of Java code.

Thanks for your help.
Alberto.

Paul Jones

unread,
Nov 17, 2009, 11:39:54 AM11/17/09
to apml-...@googlegroups.com
Hi Alberto,

The source code follows the latest published versions of the standard. I don't believe any libraries currently work with the pre-1.0 style specs (since these are not yet particularly agreed).

Have you had a look at the code in:
  http://code.google.com/p/apml-library/source/browse/#svn/trunk
?

There are a whole bunch of libraries supported there - Tim Schultz contributed a java library quite a while back, so hopefully that should be able to help you?

Thanks,
Paul.


--

You received this message because you are subscribed to the Google Groups "APML.Public.General" group.
To post to this group, send email to apml-...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/apml-public?hl=.



Alberto Bermejo

unread,
Nov 18, 2009, 7:55:26 AM11/18/09
to APML.Public.General
Thank you very much, has been a great help.

On 17 nov, 17:39, Paul Jones <pauljone...@gmail.com> wrote:
> Hi Alberto,
>
> The source code follows the latest published versions of the standard. I
> don't believe any libraries currently work with the pre-1.0 style specs
> (since these are not yet particularly agreed).
>
> Have you had a look at the code in:
>  http://code.google.com/p/apml-library/source/browse/#svn/trunk
> ?
>
> There are a whole bunch of libraries supported there - Tim Schultz
> contributed a java library quite a while back, so hopefully that should be
> able to help you?
>
> Thanks,
> Paul.
>

Mason Lee

unread,
Nov 21, 2009, 7:52:44 PM11/21/09
to apml-...@googlegroups.com
Dear Chris Saad, Paul Jones, Virginie De Bel Air,

(And to others on this list, I think Paul Jones will be writing up some summary and proposed spec mods soon, where the context here might then make more sense.)

In a recent meeting in San Francisco we talked about an updated APML spec that simplifies the interest ontology (person/location/concept/source). For the purpose of the discussion, we had talked about reducing these to a single "apml:concept" element:

<apml:concept key="[xs:string]" value="[-1.0 - 1.0]" .../>

I'd like to offer two syntactic alternatives, attempting to be more precise with the terminology used:

Suggestion 1:

<ampl:interest key="[xs:string]" value="[-1.0 - 1.0]" .../>

Suggestion 2:

<apml:attention concept="[xs:string]" interest="[-1.0 - 1.0]" .../>


Hope these thoughts are helpful. Looking forward to the new stuff.


Cheers,

Mason
http://masonlee.org

Paul Jones

unread,
Nov 21, 2009, 9:46:11 PM11/21/09
to apml-...@googlegroups.com
Mason,

Thanks for the thoughts - I think it would certainly be an idea to change the name of the element. I'd personally prefer to go down the path of suggestion 1 (interest), as opposed to an "attention" element, which could get rather repetitive.


Everyone wondering what the context of Mason's remarks are:

At a meetup between a few interested APML parties in San Francisco, we discussed some changes to get APML toward 1.0. The primary topic of discussion was to migrate APML from being a standalone document to an Atom-embeddable content. The move toward a stream-based approach has a number of benefits:
- It makes it exceedingly clear that attention is an entity that changes over time;
- The updated and from attributes are no longer necessary, because these are inferrable from more standard Atom tags (such as id, published, updated);
- The container elements are no longer required, eliminating potential noise from the specification;

On top of this, it was widely felt that simplifying down to a single XML element (with a name under discussion below) with a "type" attribute indicating the ontology to which the attribute belongs. The element would also have (as per discussions on this previously on list), an optional uri attribute. A natural language text element would always be the required subject of the element, but the URI would make it possible for source systems with a richer ontology to support providing this information in a richer form.

The general theme of these changes is that APML should be as easy as possible to create; and consumers should be able operate at various levels of intelligence. Simple consumers will be able to make a number of assumptions about the stream that may lead to their data to be slightly inaccurate (but not so much as to be useless); whilst more complex consumers will be able to track the data in a more rigid way, and thus derive more from the stream. It is generally agreed that in terms of where effort should be expent, it is far better for a consumer to do slightly more work, since the consumer will always be the application with the most to gain from APML compatibility (whereas a producer has far less motivation to provide the information if it is difficult to construct).

A stream-based approach certainly has its challenges - but the synergies with other stream based data sources present an opportunity too good to ignore. A number of details for interpreting the stream have been considered and discussed, but I'd prefer to leave these details for the updated specification I plan to publish in the next week or so (if only to ensure that I can phrase them in a clear enough manner).

Thanks,
Paul.

--

You received this message because you are subscribed to the Google Groups "APML.Public.General" group.
To post to this group, send email to apml-...@googlegroups.com.
To unsubscribe from this group, send email to apml-public...@googlegroups.com.

Mason Lee

unread,
Nov 21, 2009, 11:04:28 PM11/21/09
to apml-...@googlegroups.com
Hi Paul,

<apml:interest key="..." value="..." .../> was the one that made most sense to me as well in light of the example uses we were discussing.

Does that leave open the question of how we might express general attention? Is there a difference between interest and attention? Exploring that quickly:

Dictionary definitions:
Attention - notice taken of someone or something.
Interest - the state of wanting to know or learn about something or someone.

I pay a lot of attention to the road when I'm driving, but truth be told, I'm not that interested in roads. Yet, it is worth being able to express that I have paid attention to roads-- I know they exist- they are part of my personal ontology. Conversely, Chris Saad has a good example: He is very interested in an upcoming movie, but since it's not out yet and few people are talking about it, is rarely paying attention to it for the time being.

My browser would be good at publishing my "attention", and Netflix (with its ratings) would be good at publishing my "interest".

One thing we talked about was that the degree of attention, irrespective of interest level, might be expressible in a stream simply by the frequency of <apml:interest .../> occurrences. I wonder if that's good enough, though-- it does require looking at a history and can't be inferred from a snapshot.

What if they were combined?

<apml:attention key="..." value="1.0" interest="1.0" .../>, where value is the degree of attention, and interest is the degree of interest. (perhaps value defaults to 1.0, interest defaults to 0.0)

Cheers,
Mason
--
http://masonlee.org

Gérard Dupont

unread,
Nov 22, 2009, 5:36:09 AM11/22/09
to apml-...@googlegroups.com
Hi,

I like the idea and it has not been proposed in APML list so far. I
think it worst to have in next iteration of the model in order to
distinguish attention from interests. Indeed from browser point of
view, we can only gather "attention" data and then it's up to the
system to infer interests (or with user collaboration).

Well, nice view !

cheers
Gérard Dupont
Information Processing Control and Cognition (IPCC) - EADS DS
http://forge.ow2.org/projects/weblab/

Document & Learning team - LITIS Laboratory

Gérard Dupont

unread,
Nov 22, 2009, 5:44:31 AM11/22/09
to apml-...@googlegroups.com
Catching up on the previous messages (unfortunately I'm not on the right side of earth and thus I was not in the last APML meeting in SF).

The stream view of APML does make a lot of sense since it really describe moving attention over the time and then interests could be inferred in a way out of that cumulating attention on similar items over the past for instance. Technically I rather in favor of RSS instead of Atom (but it's probably not fixed) due to its broad used and some possible semantic extension.

Well, is there any draft of new spec ?

2009/11/22 Gérard Dupont <ger.d...@gmail.com>

thieme

unread,
Nov 23, 2009, 5:31:34 AM11/23/09
to apml-...@googlegroups.com
Hi all,
with continuous interest I am following your conversation. 

I would like to take the opportunity to let you know I am currently trying to get grips of someting related to attention: "value". Rather than focusing on interest, I would like to understand how a sense of value can be derived from interactions with content (or people). The two challenged are:
  • How to model value with user interactions (counting views? ratings? links pointing toward the uri? etc.)
  • How to model context (since value is always embedded in a context)? Which factors should be integrated in the interaction: the tags assigned to the resource? The (profile of the) person doing the interaction? The resource pointing toward, or integrating the resource?
More than just interest, which might be easily modeled, I would like to know how the above, hence: a profile representing context and value parameters, can be modeled? In a closed environment, one could think of: tags and categories representing the context, and views and ratings representing value. Maybe you would get something like 

<qpml:quality key="..." value="..." .../

If anyone is interested in discussing this with me, or has ideas about this, please contact me. I do not want to mix this conversation with the APML conversation, even though they are very much related.

Bye,

Thieme Hennis
PhD student Systems Engineering, Delft Univ. of Technology

2009/11/22 Gérard Dupont <ger.d...@gmail.com>

Gérard Dupont

unread,
Nov 23, 2009, 5:49:37 AM11/23/09
to apml-...@googlegroups.com
Hi Hennis and welcome on the list,

First you should understand that APML is just a format and thus we do not intend to find solution to all the algorithmic question you have. For sure those are really interesting and I'm pretty sure that all of us have these in mind but it's really research oriented. On the other hand APML is more oriented to data interoperability and thus finding a technical solution to exchange information related to attention and interests.

However I think the context problem you mentionned is part of the scope of the format. In the current spec it is more or less described by the sources and date. It's quite light and does not embed everything but simple enough to be usefull. We are now thinking to integrate more semantic in the model and thus provide context to concept using semantic description and links. I think your view of context is broader than this. You could start a new discussion and make a proposal of your suggestion (plane example are quite effective to describe your idea).

cheers

Paul Jones

unread,
Nov 23, 2009, 2:20:33 PM11/23/09
to apml-...@googlegroups.com
Hi Mason,

On Sun, Nov 22, 2009 at 4:04 AM, Mason Lee <maso...@gmail.com> wrote:
 
<apml:interest key="..." value="..." .../> was the one that made most sense to me as well in light of the example uses we were discussing.

 
Great. 

Does that leave open the question of how we might express general attention?  Is there a difference between interest and attention?  Exploring that quickly:

Dictionary definitions:
Attention - notice taken of someone or something.
Interest - the state of wanting to know or learn about something or someone.

I pay a lot of attention to the road when I'm driving, but truth be told, I'm not that interested in roads.  Yet, it is worth being able to express that I have paid attention to roads-- I know they exist- they are part of my personal ontology.  Conversely, Chris Saad has a good example: He is very interested in an upcoming movie, but since it's not out yet and few people are talking about it, is rarely paying attention to it for the time being.

My browser would be good at publishing my "attention", and Netflix (with its ratings) would be good at publishing my "interest".

One thing we talked about was that the degree of attention, irrespective of interest level, might be expressible in a stream simply by the frequency of <apml:interest .../> occurrences.  I wonder if that's good enough, though-- it does require looking at a history and can't be inferred from a snapshot.

What if they were combined?

<apml:attention key="..." value="1.0" interest="1.0" .../>, where value is the degree of attention, and interest is the degree of interest.  (perhaps value defaults to 1.0, interest defaults to 0.0)


I somewhat worry that having multiple values here might make things more complicated. My take on this this is that essentially you're wanting a vector in 2D space here - kind of like a current value along with a velocity to get there. Gathering this data could also be quite hard in quite a lot of cases - since attention is probably very easily confused with interest (in your road example, from an external perspective you could be as you say, not interested; or you could really enjoy your roads, and drive in your car specifically to drive them) - gathering this from the actions you perform isn't particularly easy.

Paul.

Paul Jones

unread,
Nov 23, 2009, 2:22:37 PM11/23/09
to apml-...@googlegroups.com
Hi Gerard,

2009/11/22 Gérard Dupont <ger.d...@gmail.com>

The stream view of APML does make a lot of sense since it really describe moving attention over the time and then interests could be inferred in a way out of that cumulating attention on similar items over the past for instance. Technically I rather in favor of RSS instead of Atom (but it's probably not fixed) due to its broad used and some possible semantic extension.

I don't think the choice of containing format really changes these possibilities. It was my intention to indicate this spec could be used somewhat identically in either format. The only difficulty with doing that means that you require every participant to be able to parse both formats...
 
Well, is there any draft of new spec ?

As I said in my initial email, I'm hoping to get that out sometime in the next week.

Paul.
Reply all
Reply to author
Forward
0 new messages