[Clipart] Is anyone working on categorizing the existing images?
Karl Ove Hufthammer
karl at huftis.org
Wed Jun 9 11:21:03 PDT 2004
"Áki G. Karlsson" <aki at akademia.is> wrote in
news:opr89988gcbnhb89 at localhost:
> Can't we just have both?
> I mean, a limited number of strong categories,
Or strong keywords. In other words, a *controlled* vocabulary of
keywords (containg e.g. 'France, food and flags'). Just using
free-form keywords will lead to lots of different keywords that
mean ~the same thing (cf. IMDb). A controlled vocabulary avoids
this. Trying to create only a fixed hierarchy, and shoe-horning
each image into one category is bound to fail.
In addition to the controlled vocabulary of keywords, we can have
a hierarchy structure based on these strong keywords.
The point is: Images aren't placed in hierarchical categories,
they have keywords assigned to them. The hierarchy is a separate
The hierarchy could look like this:
Fruit and vegetables
Images and the hierarchy are only linked by the strong keywords.
Different projects may create different hierarchies for their own
needs (e.g. only a subtree, or more or less detailed (e.g.
seperate fruit and vegetables, or lump all foods in one
Of course, all these *needs* to be localisable (translatable). It
will be best if US English is *not* a 'special' language here. The
strong keywords must internally only be stored as codes (numbers
or such), and translations (English, French, &c.) linked to these.
> selection at file upload just to place the images on the
> website index with "miscellaneous" as default (assuming that
> there will be a browseable index with thumbnails somewhere)
> and weak categorization as a recommendation for keywords to be
> put into the metadata of the file itself.
I like this. The weak keywords / descriptions would usually only
apply one or perhaps two images.
But is this metadata stored in the SVG files localisable?
Karl Ove Hufthammer
More information about the clipart