Five minutes with... Reuters

Jacquie Bran, project manager with the Knowledge Management events team, speaks to Chris Charlton, principal consultant at Reuters, about his involvement implementing the company’s enterprise-wide taxonomy.

When and why did you first consider implementing a taxonomy?

Three of four years ago, when we first contemplated the replacement of our existing diverse editorial systems with a single, centralised content-management system, the need for a comprehensive taxonomy strategy soon became obvious from the conceptual designs and architectures. You could even say that some of the problems we were attempting to resolve with the proposed new system were those caused by its absence in the first place.

What specific business needs were you aiming to fulfil?

On the creation side, we were addressing search, re-use and re-purposing. On the distribution side, we were trying to address the definition of our products, and on the client side we were putting the infrastructure in place to enable far more intuitive and comprehensive search and navigation.

What have you done to encourage and promote knowledge sharing, and what barriers have you faced?

We have implemented a very active cross-organisational working group of some very bright people openly sharing the kind of issues and problems they are facing. This is leading to the development of a new taxonomy with very broad acceptance, and the assurance that it is addressing the needs of the organisation as a whole, rather than just particular departmental needs.

How did you progress to implementing an infrastructure to support a taxonomy system, and what changes were necessary to ensure its success?

The issue of the taxonomy and the creation of the technology to implement it were very closely linked in our case – indivisible even. It was recognised that appropriate and consistent tagging of our content could only take place close to or at the point of content creation, and as such the whole development was lead from the major editorial content-management renewals programme already under way.

How did you arrive at a decision with regards to implementing a manual versus automated taxonomy system?

We do both. Automation is a very useful tool (when applied intelligently) as a timesaver for our journalists, and in the new system we routinely automate code for them (achieving accuracy typically in excess of 80 per cent). However, Reuters’s principals around trust and accuracy of our content mean that a manual review of codes is also required. In this combined way we get consistent accuracy, and I would argue that this is at least as important as what codes are applied. The sheer volume of content we process a day – 11,000 of our own headlines together with aggregation of up to a further 30,000 from third parties – also supports the need for a sensible level of automation.

How useful has your taxonomy been in retrospect?

It is still early days. We are only now having real content coded to the new taxonomy, and hence the capability of building new products and services based on that. However, signs are good, if for no other reason than we are looking at capabilities now that would have been difficult, if not impossible, before.

How have you incorporated a provision for your taxonomy to be updated and maintained?

Our work around taxonomy has not stopped at structure; we have also done considerable work around usage of the taxonomy. As well as how the taxonomy should be applied (including relevance ranking, hierarchies and related terms), considerable thought has gone into how the taxonomy should be maintained, including version management (which includes dates), and the implications of reviewing content tagged against previous versions of the taxonomy.

What are the main lessons learnt and are there any new milestones on the horizon?

We have learnt a lot from this work. I think we had already recognised the scale of work required, but the complexity and detail has quite surprised some people. I think it stresses the importance of taking a strategic approach rather than trying to retro-fit aspects of taxonomy over time. Another issue is the separation of the highly technical details of our taxonomy and its use with the actual technology itself. The application of the taxonomy proves far easier if the technical details of usage and such like have already been worked out.

In terms of milestones, at the moment we are concentrating on two things: extending the usage of the taxonomy internally (and hence deriving the benefit from our investment so far), and then leveraging aspects of the taxonomy for our customers – and providing value and reducing our mutual cost bases as a result.

Chris Charlton is principal consultant at Reuters. He can be contacted at chris.charlton@reuters.com