[topicmapmail] Two Models of Facets

[Murray Altheim]

Thomas B. Passin wrote:
> Murray, this is very interesting and I am glad you have broughtit up.

Thanks -- it's probably long overdue.

> Murray Altheim wrote:
> 
> 
>>Two Models of Facets
>>--------------------
>>
>>I've been investigating various ideas about facets, 'facets' defined
>>in the general sense. 
>>
>>... Since we've begun thinking about Topic Maps as structures for repre-
>>senting knowledge and knowledge structures, as well as their original
>>intent of assisting in the classification of content, it seems a good
>>time to tackle how facets and property-value relationships might fit
>>into the Topic Map architecture, particularly in XTM (since that's
>>our currently most popular implementation). I've begun working up an
>>API for TM4J for Topic Map-based inferencing, and a lot of these
>>issues pop up pretty quickly.
>>
> 
> 
>>Model A: The Metadata Model
>>---------------------------
>>
>>The first model of facets might be considered as the RDF model,
>>where we're essentially adding property-value pairs to a resource.
> 
> As probably everyone on this list knows, I consider occurrences to be 
> well suited for property-value pairs (at least, those whose occurrence 
> data does not point to another topic).  To my mind, the question is what 
> is the difference, if any, between facets and other properties (or are 
> they properties at all?).

I think the extant literature would *generally* consider that there are
two "directions" to facets. One is used in the formulation of subjects,
the other as properties of subjects. These are probably just two angles
on the same thing. I think of facets as *perhaps* properties when used
in the context of formulating a subject, i.e., as a special instance
of the more general term "property". But they might be the same thing.
I'd say that yes, they are some form of property, if properties can be
adduced to being attached to abstract concepts (as opposed to entities).
This goes back to my statements about being either in the map/ontology
vs. in the territory/on the resource.

>>Model T: The Faceted Classification Model
>>-----------------------------------------
> 
> 
> This is certainly the model that I think of when I hear the term 
> "faceted classification".
> 
> I would like Murray to make one thing clear - are you suggesting that 
> the topic map model itself be changed to accomodate facets, or are you 
> suggesting that some convention or standard or PSI set be developed to 
> provide a standard way to model facets within the exising topic map 
> model?  I do not myself see any need to change the TM model, and so I 
> take it that Murray means the second alternative - is that right?

I don't think the Topic Map model should be changed to accommodate facets,
no. But I do think that it's quite conceivable that the formulation of
subject identity via the use of facets or Faceted Classification could
easily sit as a layer above a Topic Map model, as one of many potential
models used in the formation of subject or subject identity. We've seen
a few competing ones, and even thrown out one (based on name identity).

> I think that my naive start at using facets would be to establish role 
> types for each facet type, all descending from a generalized facet role. 
>   Then the topic to be classified would play a role in a "facet" 
> association, to which the desired roles would be attached.

Both I and Kal have played with creating those PSI sets. It is just a
specialized association type, like occurrences. Ddepending on which Model
A or T we're talking about, there's some differences, such as whether
we're creating facets for our resources or facets for our subjects/Topics,
i.e., the RDF vs. FC models (Models A and T).

> As a matter of practical performance when retrieving by facet 
> collections, this scheme would probably be miserable for a large map 
> using a general purpose engine, so I would probably look to create 
> customized data structures and methods for handling the facets.  I would 
> like the topic types for the roles and the facet association to be PSIs, 
> so that my engine would be able to recognize when to use the custom 
> features.  This would not prevent me from serializing out to XTM, since 
> it would just be application-specific customization.

I don't see this as any more a performance hog than any other scheme of
similar power. If I want "all Topics playing the role of 'blue' in a
'hasEyeColor' association", where that's implemented in the model won't
have a huge impact on the amount of information modelled. And my guess
is that simply using an FC model would have some benefits for understanding
the structure of the ontology, just as people now use slots rather than
putting everything at the same level of the model.

> To me, the important thing about facets is not so much that a subject 
> can be classified by many facets, but that each facet can have its own 
> classification hierarchy (or perhaps pseudo-hierarchy).  How to make 
> that happen and to attach them to the classified subject are important 
> things to be worked out, it seems to me.  According to what I describe d 
> as my naive view, the head of each facet hierarchy could be in the facet 
> role types.  That would establish what kind of facet the role would be 
> representing.  Should the actual role-players be members of the 
> hierarchy?  That would be my first thought.
> 
> Does this make some kind of sense?

Yes, certainly, though I'd be careful to think that the components of
that facet hierarchy are somehow different from the rest of the Topic
Map ontology. Anything can serve as a facet, anything can be faceted.
Heck, I can imagine that a facet might serve as a facet type for itself.

As to the method of attachment, in XTM it'd just be a specialized
association type, in the same way we now understand that occurrences
are just a specialized association type. It's just that in XTM we didn't
create a syntax to support facets. We could have, but we don't really
need to. It might be a bit easier syntactically if we had, just as
it is easier to add an <occurrence> to a <topic> rather than creating
a new <association>, we could have created a <facet> element. I'm sure
there are those who would like that. But we'd first have to nail down
*one* definition of facet, which is kinda what I'm trying to do here.
My suspicion is that there are two different ones, one for resources
(Model A) and one for Topics (Model T).

Murray

...........................................................................
Murray Altheim                         http://kmi.open.ac.uk/people/murray/
Knowledge Media Institute
The Open University, Milton Keynes, Bucks, MK7 6AA, UK                    .

   Q: So exactly how is Ahmad Chalabi different from Manuel Noriega?

      http://www.nationalreview.com/comment/comment-singer062002.asp
      http://www.iraqinews.com/people_chalabi.shtml
      http://truthout.org/docs_03/041103F.shtml

   A: One speaks fluent Arabic, the other Spanish.

     "Noriega took refuge in the Vatican embassy, where US troops played
      hard rock music until Noriega surrendered on January 3, 1990."
      http://www.nationmaster.com/encyclopedia/Manuel-Noriega
      http://www.gwu.edu/~nsarchiv/NSAEBB/NSAEBB2/nsaebb2.htm#3a
      http://www.addictedtowar.com/panama.htm