jump to navigation

Getty Catalyst image search May 20, 2008

Posted by Mia in Frontiers.
Tags: , ,
add a comment

Getty’s Catalyst image search site is an impressive advance, allowing users to manipulate concepts in the information space by using drag-and-drop.

The Catalyst search demonstrates a very effective use of tag clouds to add/subtract terms as filters. One of its strengths is that it works by recognition, rather than by articulation. There is an offering of terms at the left, some of which you may ‘recognize’ as useful–rather than having to articulate a priori (i.e., by first dreaming up and them typing them into an “advanced boolean search” box) terms which may not be in the dataset anyway.  

I’m especially inclined to think that the insights behind this search functionality were informed by the fact that images are ‘text-free zones’ so the Getty designers/developers are likely much more accustomed to exercising their right-brain functionality. These are the kinds of search models that need to be explored and ported to left-brain dominant activities, like reading text on the screen. Far too many interfaces of our search interfaces can’t find their way off the flat, two-dimensional reading plane — but Getty has.

Only the initial search is needed to hit the ’side of the barn’ (see Bates).

Subject terms are insufficient May 3, 2008

Posted by Mia in Frontiers, cataloguing, metadata.
Tags: ,
add a comment

There is plenty of evidence that users will respond to concept clusters that represent something other than the subject of the thing: and by that I mean the genre of the thing. Although it may be a term that is somewhat oblique, genre represents facets that are based on the functionality, purpose, presentation, or use of the resource.

Humans have commonly-held concepts that turn out to be recognizable groupings of things which aren’t subjects, that is, they are groupings which are not concerned with the aboutness of the thing.

The commonly-held and shared conception of the utility of a ‘telephone book’ does not include reading such a thing from beginning to end, since this type of collection of words/numbers isn’t intended to be used in such a fashion. That doesn’t mean that a telephone book cannot be used in that fashion (i.e., read front to back) to alleviate wakefulness, or for the purposes of a filibuster, and so on, but those are instances of atypical uses.

User domain is important, though. There can be multiple commonly-shared views of the thing, e.g., telephone books can be recognizable members of other genres, for example “booster seats” or “height extenders.” Different purpose, different genre, different domain.

We should be increasingly concerned to expose genre, since it allows the user to focus on things that groups of words do other than be read narratively as strings.

Particularly when there is a surfeit of words (for example, massive retrieval results), the way those words are grouped or presented, their purpose and functionality, rises to the forefront and provides an extremely useful handle. The grouping of words is fundamentally how language is acquired in the first place, so it’s particularly important to make note of that resonance.

Genre can be a superior filter because it can more closely associate the intention of the searcher/viewer with a group of found information objects.

Tag clouds are cloudy May 3, 2008

Posted by Mia in FRBR, Frontiers, metadata.
Tags: ,
add a comment

I don’t know how the phrase ‘tag cloud’ originated, and I really don’t have anything against it except that, well, they are word frequencies clustered into clouds. It might be neat to explore word rivers, or word wells, or word mountains, or — wait, I’m having flashbacks to the dada movement and concrete poetry — so let’s leave our heads in the clouds for a moment.

Clouds aren’t good things when we want to make clear conceptual distinctions. When we want to clarify something, and help someone navigate an information space, we want to make an effort to dispel ambiguities. We want murkiness (and cloudiness) to be replaced by clarity of thought and precision.

Sometimes we really do want things to be fuzzy and ill-defined. And if that is the case, then a cloudy kind of randomness might be acceptable, fun, interesting. More art, less retrieval. Randomness doesn’t scale, though, in terms of navigation or findability or retrievability. For me that kind of randomness is like the eclectic mix of a rummage or garage sale, which are admittedly popular things, but that kind of fragmentation is a time waster for me. Experimentation on a grand scale like the Photosynth demo (TED talk) a while back by Blaise Aguera y Arcas is super interesting, though, but a wee bit out of reach at the moment.

I emphatically support the idea of user-generated tags to augment controlled vocabulary files and to greatly increase lead-in vocabulary. It would not be difficult to differentiate user tags from authorized headings in terms of presentation (different font, size, color, etc.) Bates wrote of the need for an end-user thesaurus quite some time ago, which does not appear to have gotten much traction (and which just shows how much ahead of her time her research always is).

In contradistinction to clouds, the current work happening on earth right now in the Topic Maps community (e.g., Danenbarger; Oh; others) seems to me to be on target and gratifying.

As for the ultimate utility and longevity of tag clouds, especially given the abundance of long, well-established and justifiable human-scale metaphors associated with this meteorological phenomena?

Haven’t the foggiest ;-)