Rob Stein from Indianapolis Museum of Art asked me on the STEVE list –
Do you have a feel[ing] for how many people are actually entering the collection through the tag cloud you have on your page versus how many are using the category listings? I’ve often wondered if the nature of a tag cloud naturally bias’ big terms to get bigger, and smaller terms to disappear. Presenting a cloud like this side-by-side with the categorical hierarchy seems like an interesting comparison.
Since launch we’ve had nearly 1500 user classifications. Interestingly there seems to be no immediate pattern in the way in which objects are user classified and the rationale for classification is unsurprisingly very mixed (as is our collection). Most of the larger user classifications such as ‘bowling club‘ is the result of a single user classifying multiple objects in one go. (Bowling club were all tagged on the same day and none added since).
Whilst we don’t specifically track category listing use we do track tag cloud use. Here’s the figures for the last 7 days.
Date | Total successful searches | Subset of searches using tag cloud
13/08/2006 (11,665) (4,006)
12/08/2006 (12,165) (4,847)
11/08/2006 (13,613) (1,352)
10/08/2006 (5,572) (569)
09/08/2006 (6,782) (318)
08/08/2006 (4,530) (564)
07/08/2006 (9,605) (1,638)
At its lowest tag cloud searches represent 4.68% of searches, and its highest 39.84%. That is a pretty large difference but I have a feeling that the reason for the recent few days generating both more total searches and a higher percentage of tag cloud searches is that Google has again spidered the site and picks up the tag cloud words as keywords.
Because tag cloud words are user-generated there is a greater chance that they will be ‘more used’ than words from our official taxonomies. This means not only will they be more used on the site, but that they are probably also going to be words that are more often searched for in Google as well.
Now when a user clicks a tag cloud word they get TWO sets of search results. The first set of results is a simple tag search, the second is a general free text search for that keyword.
Rather than necessarily biasing the ‘tagged’ objects what we are actually observing is that a tag cloud click more frequently results in the viewing of an untagged object which appears in the later free text results. I’ll have to keep an eye on this and see if this trend continues as more objects are tagged.
As for the categorical hierarchies, we are seeing very little usage of them. The vast majority of users are using direct search terms or clicking the tag cloud, or, more often than not, getting to objects or search results via a Google search.
What has changed recently is that we have added ‘subject terms’. These are slightly looser taxonomic classifications which address particular ‘themes’. An example of this is the term ‘federation’ which is used to refer to object related to the period of Australian federation. These subject terms don’t describe the actual object but are related to its provenance and significance – and thus are particularly useful to high school teachers and students. A small portion of our total objects have subject terms attached currently and they tend to be those relatively recently acquired.
What I am noticing is a very marked appearance of subject terms in the search terms indicating that they are being used as navigation devices to discover ‘related’ objects. In the next week or two we will be making the subject terms much more prominent as it seems that they are perhaps more useful to the user than our broad object categories despite their limitations.