RE: Broad and flat vs hierarchical facets - beginning with introducing myself

Bill,

Your question is great.  I faced a similar problem when creating a faceted
approach to library systems navigation.  The 600's subject data fields, for
instance, have a tremendous span (Library of Congress Subject Headings
(LCSH) by various estimates include roughly 250,000 terms).  A first take on
reducing this - creating orthogonal facets - showed some improvement.  For
instance, in a geography facet, I've seen users can tolerate a fair amount
of hierarchy (and this can be taken from existing hierarchies available on
the Web), because they understand the structure (e.g., have an idea of which
countries are in a continent) of the hierarchy.  

However, in LOC class navigation - or Dewey, for that matter - this is
probably not the case, because users do not have an intuitive map of the
hierarchy.  Further so in mapping any hierarchy for facets like "content
form."  Worse still, removing the geography, personal name, other name, etc.
terms from the 600's still left an enormous span of general subject terms in
a "flat list" derived from the 650's.   

I applied various other techniques to help out.  Perhaps the most important
was fusing search with the navigation.  So a user can then execute a search,
and the navigation (or hierarchical / faceted menus) can be whittled down to
those only pertaining to the results of that search.  Typical searches yield
at least one order of magnitude reduction to this problem.  For facets with
1000 terms (fairly large facets), this yields a reduction to a more
manageable 100 terms.

Still, 100 terms is quite a lot to the casual user (i.e., unless someone is
in "deep" research mode, it doesn't provide a quick, meaningful cue of how
to navigate to refine).  There are a bunch of other approaches to then help
out...For instance, relevance ranking the facet navigation options
themselves according to their prevalence in the underlying data (or user
search execution, etc).  In the case of a search for baseball, for instance,
content form: Statistics shows up a lot, because a lot of baseball books are
written about statistics.  Popping this content attribute to the top using
ranking for the facets provides a likely cue to the end-user searching on
baseball (and for the end-user not searching on "likely" topics of interest,
the more deep research mode of expanding the list to the full ~100 content
forms in alpha may be appropriate)...

I agree with the nature of your question - faceted classification navigation
systems can present a host of new and extremely important implementation
challenges!  Hopefully you found this answer at least somewhat useful...

Best,

Matt.

-----Original Message-----
From: Roy Roebuck [mailto:[email protected]]
Sent: Wednesday, February 25, 2004 6:22 AM
To: [email protected]
Subject: RE: [facetedclassification] Broad and flat vs hierarchical
facets - beginning with introducing myself


Hello:

 

I suggest you take a look at my faceted classification approach at
http://one-world-is.com/beam.  It provides a six faceted taxonomy of
natural/human classes.  These are location, organization, organization unit,
function, process, and resource.  When the six facets/classes are arrayed
against each other as a "drill down" matrix, the combinations of associated
classes and instances provides an ontological architecture for a generalized
knowledge base.

 

The primary design intention of this ontology, which integrates what I refer
to as operation and intelligence management ontologies into a spiral life
cycle, was to provide a means for whole-enterprise management and dynamic
situational awareness for all participants in the enterprise value chains.
It can also be used as a knowledge-base to organize and manage an
organization's enterprise architecture (e.g., U.S. FEA and DoDAF/C4ISR,
OpenGroup TOGAF8).

 

 

 

Roy

 

Roy Roebuck

Ph: 703-598-2351

[email protected]

one-world-is.com/beam

Calendar: http://one-world-is.com/freebusy/roebuckr.htm

FreeBusy: http://one-world-is.com/freebusy/roebuckr.vfb

-----Original Message-----
From: HALL Bill [mailto:[email protected]] 
Sent: Wednesday, February 25, 2004 12:38 AM
To: [email protected]
Subject: [facetedclassification] Broad and flat vs hierarchical facets -
beginning with introducing myself

 

As someone who has recently joined this list I am seeking some
information from people who have pragmatic experience implementing
faceted classifications - but first an introduction and background.

I am currently working in the Australian defence industry as a
documentation and knowledge management systems analyst. I earned my PhD
in evolutionary biology in 1973 (where systematics was a major part of
my doctoral research). 

Throughout my academic career I had a particular bent information
science, and often explained to librarians how the various computerised
indexing systems worked. I am familiar with Ranganathan and a variety of
other indexing schemes used in library environments.

After reaching an academic dead-end around 1980 I ended up in the
computer industry as a technical writer, documentation manager,
documentation systems analyst and designer, and am currently working
with enterprise knowledge management systems - still more or less in
analysis and design roles. Since 1990 I have been working exclusively
within Tenix Defence (and its predecessor companies). Tenix is a
privately held company and is Australia's largest defence prime
contractor. We are currently finishing the $A 6 BN fixed price ANZAC
Ship Project to build 10 frigates - 8 for Australia and 2 for New
Zealand.

>From a theoretical point of view I am interested in the roles of data,
information and knowledge management technologies as extensions to human
and organizational cognition. A summary of what I have been working on
is available on the Web: Hall, W.P. (2003). Working Towards Biologically
Based Theories of Organization and Knowledge to Understand How
Organizations Work Best. Joint Lunchtime Seminar, 08/09/03, Rutherford
House, Wellington. New Zealand Knowledge Management Network and
Government Information Systems Forum.
[http://www.nzkm.net/assets/NZKMNetGovis(1).pdf]
<http://www.nzkm.net/assets/NZKMNetGovis(1).pdf%5d> 

Which brings me to what I am doing at the moment.

We have implemented OpenText's LiveLink application as a corporate
intranet portal, where I am trying to implement a classification
structure that engineers, office staff, project managers, commercial
people and the like can apply to objects they place in the system
(documents, people's contact details and specialities, registration and
location of paper files, etc.), and that can then be used by the same
people to intelligently retrieve information from the system.

I'm convinced that no one dimensional classification system can cope -
especially where people applying the classification metadata are not
skilled records managers. Facets seem to be the way to go - and appear
to be well within the capabilities of our platform to apply and search
by in a user-friendly way using LiveLink's "categories" and/or the
"classification" addin module.

Our current prototype includes 7 independent facets - where we expect
users to apply only those facets that are relevant to a particular
object being classified: 

o  Business Unit (object relating to a particular business unit), 
o  Other Organisation (e.g., document source, party in a contract, etc.)
o  Project Name/ID (most of what we do relates to an internal or
external project)
o  Project Lifecycle Phase (we are implementing a project lifecycle
management methodology based on PMBoK so this should be meaningful to
many of our users)
o  Subject Matter (e.g., type of content from a functional point of
view)
o  Document/Object Type (in principle our system can point to a person
as well as it can a document)
o  Discipline (e.g., legal, engineering, H/R, etc.).

The intention is that all selections of identifiers in each facet either
to apply an object or to search will be made from picklists (these are
also easy to create and maintain). My problem is that some of the lists
(e.g., projects, other organizations) may have thousands of items. My
hope (yet to be fully verified) is that we can implement the
classifications module such that choices can be made in a hierarchical
fashion. By limiting search to the Boolean intersection of several
facets - retrieval can be quite precise even where reasonably general
terms are used.

My questions to the list is for pragmatic experience re 

1. What is the practical maximum number of items that can be included in
a flat (alpha or logically sorted) picklist before users won't use it or
can't find the item they are looking for?

2. In a hierarchical picklist, how many layers are considered to be
practical and what is the practical maximum number of items that should
be presented in a layer (assuming that the hierarchical logic allows
much freedom to go flat vs deep)? 

Other experience-based comments based on working with systems like I
have described would be welcome.

Many thanks,

Bill Hall

Documentation Systems Analyst
Head Office, Engineering
Tenix Defense
Williamstown, Vic. 3016
Phone: 03 9244 4820
Email:[email protected]
URL: http://www.tenix.com

Honorary Research Fellow
Knowledge Management Lab
School of Information Management & Systems
Monash University
Caulfield East, Vic. 3145
URL: http://www.sims.monash.edu.au/research/km/



----------
Thanks for playing. 




  _____  

Yahoo! Groups Links

*         To visit your group on the web, go to:
http://groups.yahoo.com/group/facetedclassification/
  

*         To unsubscribe from this group, send an email to:
[email protected]
<mailto:[email protected]?subject=Unsubscrib
e> 
  

*         Your use of Yahoo! Groups is subject to the Yahoo! Terms of
<http://docs.yahoo.com/info/terms/>  Service. 



[Non-text portions of this message have been removed]



----------
Thanks for playing. 
Yahoo! Groups Links



 


------------------------ Yahoo! Groups Sponsor ---------------------~-->
Buy Ink Cartridges or Refill Kits for your HP, Epson, Canon or Lexmark
Printer at MyInks.com. Free s/h on orders $50 or more to the US & Canada.
http://www.c1tracking.com/l.asp?cid=5511
http://us.click.yahoo.com/mOAaAA/3exGAA/qnsNAA/0bmwlB/TM
---------------------------------------------------------------------~->

----------
Thanks for playing. 
Yahoo! Groups Links

<*> To visit your group on the web, go to:
     http://groups.yahoo.com/group/facetedclassification/

<*> To unsubscribe from this group, send an email to:
     [email protected]

<*> Your use of Yahoo! Groups is subject to:
     http://docs.yahoo.com/info/terms/