3 Object
An object is characterized as any entity that may be retrieved
and/or given descriptors by users. Each object is uniquely identified
by an oid (object identifier), and associated to a logical
description (see Section 2).
Most often objects represent entities out of the context, mainly
files, web pages, or parts of them (e.g., pictures, BibTeX entries,
emails, URLs, songs). This makes it possible to automatically extract
object descriptors from the content of these external entities (e.g.,
MP3 tags, sender and subject of emails, size of pictures). However not
every descriptor can be automatically extracted, and users can add
their own descriptors to objects (e.g., priority of an email, ranking
of songs, event associated to a picture). The former descriptors are
said intrinsic, while the latter are said extrinsic. Both
kinds of descriptors are important as intrinsic descriptors save a lot
of time to users, and extrinsic descriptors allows an arbitrary
personalisation of descriptions.
From each object description a set of features is automatically
extracted. Each extracted feature subsumes the description and
represents a more or less general aspect of the object (e.g., a title
word, the genre of a song, a size interval). Without these features
the classification of objects would be completely flat as object
descriptions are often unique and incomparable. They play a
central role in browsing a context, which is presented in the next
section.