I've been thinking about how topics are organized in
K-Collector.
As of my writing the largest K-Collector site is
W4 which currently has
726 topics defined in 3 classifications.
- What - 410 topics
- Who - 227 topics
- Where - 89 topics
This is beginning to present some challenges. For example current
interfaces in K-Collector clients for selecting topics highlite
suggested topics but also display all available topics (under their
classifications) for you to choose from. The problem is that,
with this number of topics, it's all rather unwieldy.
I already think that it would be advantageous to allow a further
subdivision of the classifications to form a 3-level hierarchy.
For example What could subdivide into things like products, protoocols,
principles, patterns and so on (sort of modelled after Denham Grays
Information Gathering Template). But this is not a complete answer.
First I do not want to grow arbitrarily nested taxonomies. If you
think about your own experience with menu bars, how often do you look
more than 3 levels deep? And how irritating is it to have to poke
around like that? Also the deeper the taxonomy the more effort
has to be put into designing it and this is the domain of experts and
to be avoided
Second faceted classifications only work when they narrow things.
As you descend the hierarchy you become more precise about the term you
are talking about (Anything => any person => a member of a group
=> an individual). But for example, when I am writing a post
about politics and economics I am really only interested in topics
related to those subjects, i.e. I have cross-cutting concerns (like
Aspects w.r.t. OOP)
I've been thinking a little bit about Peter Van Dijck's
suggestion for
nested facets. As Travis Wilson described it in a post to the facetedclassification list:
For example, Peter's article poses a "Product Type" facet where "Cameras"
is a heading down in the taxonomy somewhere. Cameras have certain
properties like "Resolution" and "Lens Type" that just aren't relevant to,
say, hubcaps. So "Resolution" and "Lens Type" are facets with a scope of
"Product Type = Camera". A faceted navigation interface would expose them
whenever a search was already restricted to "Camera". Otherwise, they're
structured like every other facet you've ever seen.
to see whether there is a role for nested facets. I'm still thinking about that.
It's also possible that we could also use the many relations (thousands
and thousands of them) that K-Collector builds up to create dynamic
cross-cutting hierarchies. The idea here would be to take one or
two dominant topics and then order all other topics according to how
relevant they were. This should, in theory, put more relevant
topics closer to your attention.
No solutions yet, just questions & ideas.