[sigia-l] Faceted browsing WAS The category of "Miscellaneous"

[Jonathan Broad]

On Jun 16, 2004, at 12:02 PM, Donna M. Fritzsche wrote:
>>
>> You've gone beyond "parameters/attributes of x" at that point, 
>> because the values of the parameters themselves have 
>> attributes--parents, siblings, children, or even arbitrary 
>> relationships in the case of Topic Map representations.
>
> It is always the case that the values of given 
> attributes/facets/parameters have relationships to each other. 
> Otherwise they would not be useful in describing a given object.
>

I wouldn't want to argue with that!  A flat facet's terms are all 
related--to the facet, as valid values.  And, more importantly, any 
term is related to other terms (some more, some less) in the vague 
cloud of semantic association that exists in an end-user's head.

That point may seem trivial at times, until you don't understand the 
language used in the metadata. The utility of the metadata drops 
sharply.  :)  But you missed my point, I think, which is that the 
situation is qualitatively different when a system's been designed to 
read *the relationships* as well, and merge them into search results in 
various ways.

> The fact that that structure has been pulled out and 
> identified/formalized/described in a more definite manner (controlled 
> vocab, thesaurus, ontology, semantic net, etc) is an independent topic 
> from the browsing and search utitlity of the basic attribute/value 
> structure which is part of many search/query/ and database systems.  
> The utility of the basic functionality of being able to search or 
> retrieve based on the attribute/value paradigm is not changed by the 
> introduction of a formal manner for representing the value space, 
> although it may be enhanced.

Well, I got into this thread to object to your assumption, which is 
that metadata structure is necessarily independent from actual assigned 
and searchable metadata, so here goes:

Formal vocabulary structure and keyword searching are only independent 
while the structure is not machine-readable and actionable in some way. 
  When the structure is as queryable as the individual item's metadata, 
they can be combined in novel ways.

The key difference is that at any given point in a faceted 
search/browse experience, you know the relationships your current 
result set has--both to the controlled vocabulary (displaying 
still-valid facets, or parent or child topics, for instance), and also 
to the collection as a whole (displaying the quantity of still-valid 
results for the next potential combination of facets).

The whole is greater than the sum of its parts, which is the point I'm 
trying to make--parametric search is essentially more limited than 
full-on faceted browsing, which is why I kinda (just kinda!) object to 
the term *when applied to faceted browsing*.  Except in marketing, 
which can't be helped.

Jonathan