Context-centered IR is an expression which can be used to
encompass all tools, techniques and algorithms finalized at
producing a search outcome (in response to a user’s query),
which is tailored to the specific context. This way the “one size
fits all” approach is no more valid. When context is referred to
the user context, we may talk about personalized IR.
The previous short introduction to the notion of context and
its possible use in IR makes it evident that in order to
implement a context dependent IR strategy, two main activities
must be undertaken, as sketched in Fig.1. The prerequisite
activity is of type knowledge representation, and is aimed at the
definition of the context model. Such an activity comprises
sub-activities such as the identification of the basic knowledge
which characterizes the context, the choice of a formal
language by which to represent this knowledge, and a strategy
to update this knowledge (to adapt the representation to context
variations). The second activity is aimed at defining processes
(algorithms), which, based on both the knowledge represented
in the context representation and the user query, are finalized to
produce as a search outcome an estimate of document relevance
which takes into account the context dimension(s). In other
words, the context is used to leverage the effectiveness of the
search outcome. As it will be explained in section III this can be
done by different approaches, which can be classified
depending on the way in which the contextual information is
exploited.
While in this section we have introduced a general definition
of context, and of context-centered IR, in the following sections
we will focus on personalized IR, i.e. to IR approaches which
take advantage of the knowledge represented in a user model,
also called user’s profile. ly
Context-centered IR is an expression which can be used to
encompass all tools, techniques and algorithms finalized at
producing a search outcome (in response to a user’s query),
which is tailored to the specific context. This way the “one size
fits all” approach is no more valid. When context is referred to
the user context, we may talk about personalized IR.
The previous short introduction to the notion of context and
its possible use in IR makes it evident that in order to
implement a context dependent IR strategy, two main activities
must be undertaken, as sketched in Fig.1. The prerequisite
activity is of type knowledge representation, and is aimed at the
definition of the context model. Such an activity comprises
sub-activities such as the identification of the basic knowledge
which characterizes the context, the choice of a formal
language by which to represent this knowledge, and a strategy
to update this knowledge (to adapt the representation to context
variations). The second activity is aimed at defining processes
(algorithms), which, based on both the knowledge represented
in the context representation and the user query, are finalized to
produce as a search outcome an estimate of document relevance
which takes into account the context dimension(s). In other
words, the context is used to leverage the effectiveness of the
search outcome. As it will be explained in section III this can be
done by different approaches, which can be classified
depending on the way in which the contextual information is
exploited.
While in this section we have introduced a general definition
of context, and of context-centered IR, in the following sections
we will focus on personalized IR, i.e. to IR approaches which
take advantage of the knowledge represented in a user model,
also called user’s profile. ly
การแปล กรุณารอสักครู่..