Currently preparing the second version of our website, we have come up with an uneasy problem: Categorization of our content. This is a common problem but I felt that after doing 2 consecutive focus groups with mothers, classification of the information of interest to them was quite a difficult and broad subject. One quick solution was to add-up a few cagegories and modify them later, let them tag and check from time to time which tags come the most often and rebuild the categories afterward.
After a few not-so-successful meetings, we came up with 2 successful website where we got a positive experience with categorization and navigation (it’s clearly related). Facebook does a fairly good job at asking a good amount of data from users, data that can come both from pre-defined categories and by free tags (textboxes). And Taobao.com doing a wonderful job at presenting the information through very easy-to-browse categories.
From this point onward, I did a bit of research and came up with an interesting model, which may look pretty obvious at first but which is nevertheless very interesting.
This article is well worth reading to understand the evolution of classification models following the appartion of the web.
A) Traditional classification model (ontology or taxonomy) is in most cases inadapted to large data sets common from the web. They are difficult to maintain and to make evolve and are not very easy to navigate either for users (we are all different and don’t always think of the same category for the same item). On the other hand, usability testing shows the obvious fact that users tend to prefer organized and structured content. (nobody likes to feel “lost”).
B) The (free) tagging model is the action of letting users assign “keywords” to content without any constraint. It has become very popular for articles in blogs, links in delicious, or tagging pictures in Flickr… It has come at a time when the top-down approach of categoy models was seens as unadapted to the web, disorganized in structure. Tagging, on the contrary is a bottom-up approach which let end-users categorize the content. The model is also very evolutive. As we know with “tag-clouds” it shows the current situation of the content inside a website.
But Tagging, even though it solves many problem of traditional taxonomy (static categories defined a-priori) leads to new problems: in different research papers, UI (User Interface) experts shows that Tags are a bad navigation tool, as they lack “context” and logical relationship. Tag-clouds are a nice tool to view the current situation of a websites’ content, but they are very inefficient to navigate as keywords are often totally unrelated and a word without context can have many different meanings. The lack of context and the synonyms homonyms makes it also not very practical to use for users.
To enhance the usefullness of tags for end-users some website have developped tag-clusturing models. (flickr is creating groups of pictures of similar interests by grouping pictures that have some common tags in common.) while this concept is an improvement to user experience, tests shows that users still prefer ordred (classified) data rather that simple grouping of common terms.
D) Faceted classification is another not-so-new classification concept that is gaining some interest in the web, especially associated with the Tagging concept. In short the concept is based on the idea that any item has a set of orthogonal (meaning not related) properties or dimensions, and that we can classify items by giving each of them a value on these properties. For example, if we want to classify mobile phones, we can define a few dimensions (facets) as “price”, “brand”, “color”… [note that to be a true facet, one item cannot have two values in the same dimension - it cannot be a nokia & a motorola ].
From this basic concept, follows a powerful organisation and navigation system, very adapted to the web. At the difference of traditional taxonomies, multiple facets enable many starting point in the browsing or searching experience. (you can start by brand or by price…) and add a context to the tags, which make them more meaninful for users.
This concept is very well explained in the following presentation:
