Reading Time: 3 minutes

You might think that there need be no war over creating a taxonomy. When should you use a hierarchical classification, and when a set of facets? You might think it hardly worth bothering over. Yet some thesaurus experts fight to defend hierarchical taxonomies; you feel that if there were a pecking order of taxonomies (a nice idea, that), then hierarchical taxonomies would be ranked highest. If you don’t believe me, try reading a recent post from Access Innovations, Down the Rabbit Hole. In this post, Barbara Gilles, a taxonomist at Access Innovations, defends hierarchical classification over any other. What is her argument? It would appear to be that with any other taxonomy format than hierarchical, the user is pretty much down a rabbit hole, that is, lost. To prove her point, Ms Gilles uses examples from the US Department of Education’s ERIC Thesaurus, dating from 1964, and indeed shows some illogicalities when she tries to go up or down its rather flat list.


Yet the experience we all of us have when shopping on the web is that faceted search is exactly what is required. It’s not hierarchical, because the criteria we use when buying, say, a TV, are not related to each other. We may choose a 32″ screen or a 40″ screen, and we may choose a plasma display or an LED display but the choice of one is unrelated to the choice of the other (OK, plasma displays tend to be larger, but there is no intrinsic connection between technology and size in this case). The example of any major retailer website will provide examples of faceted search – I selected as a particularly good example of faceted search.




In fact, they helpfully enable you to choose comparable products by a specified facet. But at the same time, I’m sure Barbara Gilles is correct that the ERIC Thesaurus should be hierarchical. So can we draw a distinction between the two types? Fortunately, help is available on the Web. An article by William Denton from 2003, How to Make a Faceted Classification and Put It On the Web provides a clear way to decide which type to choose. Basically, use facets if the categories are mutually exclusive – Denton’s example is dishwasher detergent. Facets could be how packaged (tablet, powder, liquid), how priced (low, medium, high), and so on. In contrast, hierarchies are used where there is a relationship between all the members of the group. Apples, pears and apricots are all types of tree, or if you want to be very precise, use a Linnaean hierarchy (this is what we all think of when we imagine a taxonomy) where each new group has characteristics related to its parent groups. What is interesting, then, is not the distinction, but why some taxonomists seem determined to fight for one type over another. Sometimes the preferences of taxonomists seem more difficult to grasp than the subjects they are classifying.