Curiouser and Curiouser!

Curiouser and Curiouser! on taxonomy http://matt.blogs.it/ RSS feed for topic taxonomy Copyright 2006 Matt Mower Squib/0.1 self@mattmower.com self@mattmower.com en-gb Teenage Taxonomies Sat, 01 May 2004 10:40:23 +0100 I've been thinking about how topics are organized in <a href="http://www.evectors.com/itkcollector/">K-Collector</a>. As of my writing the largest K-Collector site is <a href="http://w4.evectors.it/">W4</a> which currently has 726 topics defined in 3 classifications. <ul> <li>What - 410 topics </li> <li>Who - 227 topics </li> <li>Where - 89 topics </li> </ul> 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 <a href="http://denham.typepad.com/km/2003/11/information_gat.html">Information Gathering Template</a>). 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 <a href="http://server2.hostvalu.com/pipermail/discuss_aosd.net/2002-January/000052.html">Aspects w.r.t. OOP</a>) I've been thinking a little bit about Peter Van Dijck's <a href="http://poorbuthappy.com/ease/archives/002822.html">suggestion</a> for nested facets. As Travis Wilson described it in a post to the facetedclassification list: <blockquote> <pre wrap="">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.</pre> </blockquote> 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. http://matt.blogs.it/entries/00002159.html Visualizing concepts in weblogs Tue, 06 Jul 2004 13:10:30 +0100 <a href="http://anjo.blogs.com/">Anjo Anjewierden</a> & <a href="http://blog.mathemagenic.com/">Lilia</a> are demonstrating a very interesting tool for visualizing conceptual relationships within weblogs.  It uses technology similar, but more advanced, to K-Collector for analyzing the content of weblog posts and drawing out concepts which can be analyzed and displayed as a network so that you can explore the relationships.  It also allows comparison of terms between different weblogs. Very intriguing! http://matt.blogs.it/entries/00002406.html DelphiGroup: Making the case for taxonomy Mon, 16 Aug 2004 11:56:57 +0100 I'm reading an excellent report from <a href="http://www.delphigroup.com/">Delphi Group</a> called Information Intelligence: Content Classification and the Enterprise Taxonomy Practice (see <a href="http://www.delphigroup.com/coverage/taxonomy.htm">here</a> for a download link). The gist of the report is that, for enterprises, search is not enough. The report makes the case that whilst search technology has improved a lot in recent years, and continues to improve, the majority of professionals still find it an unsatisfactory way to work and often spend in the region of 20% of their time searching for information. Often cited problems were constantly changing information and a lack of precision about what they were looking for. The report then makes a case that introducing taxonomy based services can significantly improve performance and save money by eating into that 20%. Whilst the report is funded by a number of companies with a vested interest in taxonomy (for example Autonomy or Verity) the case seems to be well made off the back of a credible piece of research (which is a follow-up to similar research done last year). My summary: <ul> <li>82% of users do not have access to a centralized point of search & information across information systems</li> <li>The provision of a singular navigational front end (e.g. taxonomy) and omnipresent search tool that collectively aggregate disparate content resources, can, from an end-user perspective, deliver the simple single point of access that many users strive for.</li> <li>Lack of organisation of information is the number one problem in information management & retrieval.</li> <li>If professionals are spending 20% of their time (or more) looking for information then this results in an opportunity cost & represents a runaway expense item in many organisations.</li> <li>Keyword search assumes you know what you are looking for & that it an often erroneous assumption.</li> <li>75% of people surveyed during a Yahoo market research project preferred browsing to searching.</li> <li>In some instances it is easier to discover information about a particular subject if you see it in the context of related thought. Browsing encourages associative thought.</li> <li>The availablity of taxonomy eliminates the need for the researcher to completely understand the subject before issuing a query.</li> <li>Browsing via a taxonomy in essence provides an education on the subject and lends insight into the issues or facets of the subject.</li> <li>The number one source of frustration with search of on-line content is the fact that the content they search for is constantly changing, which both frustrates the user and reduces the effectiveness of simple search.</li> <li>Use of a taxonomy can provide a dynamic bookmark so to speak, a one-stop-shopping guide to all relevant content on a subject. Return to a subject node exposes the latest and complete collection of content about that subject area. This addresses the number one cause of frustration, the dynamic, volatile nature of information sources.</li> <li>Navigation of a well-designed interface to information on a web site/portal automatically directs the researcher to other relevant topics.</li> <li>The tagging effort represents another process that a business must undertake in order to obtain the benefits of a taxonomy. In some cases this could be done manually. But this approach is not easily scalable.</li> <li>Will authors be willing or available to perform this classification manually?</li><li>68% concur that the process of locating & retrieving the information needed to effectively execute their jobs is difficult and time consuming.  Not a single respondent strongly disagreed with this statement.</li> <li>Whilst users see some improvement in information retrieval over the last 2 years, their attitude towards its level of difficulty remained virtually the same.</li> <li>Respondents overwhelmingly pointed to the fact that business content is constantly changing and has to be continually relocated.</li> <li>Findings pointed more towards the speed and ease of use of retrieval environments and less to effectiveness, as the primary point of pain amongst todays business people. </li> </ul> I think this report is very well worth reading to anyone interested in search, taxonomy, or knowledge organisation. Of course I too am baised because I think <a href="http://www.evectors.com/itkcollector/">K-Collector's</a> integrated approach addresses a number of the concerns raised by this report. http://matt.blogs.it/entries/00002449.html