Main | An alternative approach to Knowledge Management (cont’d) »

Welcome to a new blog. You may already know of me from previous writings, either in print or on IT Web sites (which might be why you’re reading this) - but if not, I am a researcher, consultant, and software developer focused on new directions in Business Process Management (BPM). I am also CTO of Role Modellers, a company whose mission is to enhance understanding and computer support of collaborative human work processes in industry.

So this blog will cover such issues. However, it will also range rather wider, and attempt to drive out the underlying trends in enterprise technology, by:

• Taking a fresh look at conventionally accepted wisdom and predictions
• Placing emerging technologies into perspective
• Providing a hands-on guide to important alternatives to mainstream technology, including up-and-coming open source projects.

If you are responsible as CTO, CIO, architect, developer or consultant for helping determine IT strategy, this blog will help you make more informed decisions.

So let’s start with a current hot topic – enterprise knowledge management. McKinsey identify “learning how to leverage knowledge” as one of their top 10 trends for 2006 - and recently I have been working with UK government scientists to help improve the ways in which they collect and analyze information about current scientific research. This has led to some interesting discoveries - both about alternative techniques to those usually employed for knowledge management (KM), and about emerging open source tools.

Since KM can be a confusing field, let’s introduce the key concepts before going further. Visit the Web site of leading KM vendors such as Autonomy or Entrieva and you will find some very high-powered, feature-rich and hugely expensive product suites, claiming to help you “understand the hidden 80%” , “make sense of unstructured information”, “empower the knowledge worker”, and so on. What does all this actually mean in practice?

In general, KM tools are focused on the "annotation" of documents with "tags". In other words, hidden fields are added inside the documents that indicate:

• The sorts of thing being discussed (categories in which the information fits)
• Relevant connections (links to other information with similar categories).

Annotations can pertain to the entire document or just parts of it – particular word phrases or multimedia objects, for example. Depending on the toolset and the customer requirements, the fields can be created manually, semi-manually, or automatically. Once a corpus of documents has been annotated, it can be viewed in various new ways to try and better understand the nature of the information, for example via a graphical depiction of the inherent categories and links.

This sounds great, and some KM vendors have been very successful. However, leading experts tend to describe the success of KM technology as “dismal” . Why?

As the interview referenced above explains, management practices lie at the heart of the problem. New technologies on their own will not improve working practices. I will have a lot more to say in future blog entries about how radical new techniques for process management can make a difference here. For now, though, let’s stay with technology, and look at an alternative to the conventional approach to KM that may well offer more power in typical business situations – and for which open source tooling is emerging.

Most KM tools make use of the RDF stack – a set of standards from the W3C that support XML annotation of documents. Each of these standards (RDF, RDF Schema and the 3 sub-languages of OWL) is based on the preceding one, and provides more expressive power at the expense of added complexity. The details are all very technical, but there are two features of the RDF stack of particular business importance.

Firstly, the lower layers of the RDF stack (the layers used by most tools) support only hierarchical categories of information: taxonomies rather than networks. This means that such tools impose an unnatural ordering upon business knowledge. For example, Autonomy claims that their: “Taxonomy Generation feature can automatically and consistently understand and create deep hierarchical contextual taxonomies of information based on conceptual understanding” and Entrieva claim that their SemioSkyline product enables “users to browse or search for information organized by category within a familiar, hierarchical structure” [my italics]. However, most business knowledge is more like a network than a tree. Knowledge categories do not generally divide neatly into parents, children, grandchildren and so on – rather, links between them may occur all over the place. Even if multiple taxonomies are used together, business people do not really see their world in hierarchical terms.

Secondly, RDF annotations are often added into the documents themselves. For knowledge associated with HTML documents the RDF Model and Syntax Specification from W3C says, "The recommended technique for embedding RDF expressions in an HTML document is simply to insert the RDF in-line" (another common approach is to use "Dublin Core" metadata within the META tags in the HEAD element of the document). If such techniques are used, as you update the documents the annotations also get updated, and it is hard to track changes to your knowledge independently of changes to your documents. Why should you need to? Well, the fundamental principle of KM is that business decisions should be based on business knowledge. Therefore, if you want to understand why a certain decision was taken - to justify it, to re-use the reasoning, or for any other reason – you need access to the knowledge that was available at the time. The knowledge available right now is no use for this purpose - but with most KM tools, that is all you get.

So, is there another approach to KM, one in some ways more in tune with real business needs? And one for which open source tools are emerging? Tune in again to this blog for more about an alternative approach to KM …

Posted by keithhb in Knowledge Management |Digg This|Add to del.icio.us

Posted by: Lucas Rodriguez Cervera at March 9, 2006 06:57 AM