Keynote

Where the Things Have no Name:
Searching in a Lexical Vacuum
by Simone Santini

Many applications of search deal with imaging data specified by well defined lexical concepts: one searches for houses, dogs, umbrellas,... Searching and finding these things, that is: transforming the lexical specification in some visual description, can be horrendously complicated but, at least, we know what we are looking for.

When we search for 3D data, at least when we do so in the typical domains in which 3D data are used, we find ourselves in a different situation. On the one hand, the models that we have are more complete, and do, in principle, allow a better match with the user desires. On the other hand, specifying exactly what a persona wants is a problem in itself.

Consider a simple example: a data base of designs of mechanical parts. Standard parts have a name, but new ones don't, and it is necessary somehow to _describe_ what we want, using suitable graphics and/or linguistic means; a task that, in itself, may be very complicated.

In some areas, like many cultural heritage areas, there is an established technical vocabulary that can be used, but in many others it is necessary to create the suitable vocabulary before we can start specifying our queries.

In this talk, we shall discuss several possible approaches to defining what we want from a data base of nameless shapes. We shall consider the integration of sketches and textual information as well as the use of linguistic games for creating a lexicon and the association to the 3D data.

In the first case, the problem is one of integration of information: somwbody may draw a sketch of a face of an object and write next to it "cylinder" or "360" to indicate that it is a rotation surface. In the scond case, the problem is to create an agreed-upon lexicon through interaction as well as the relevant connections with the visual features of the objects