Information Technology Reference
In-Depth Information
In this context, it is very useful to define the concept of relevance information. We can divide relevance
into two main classes (Harter, 1992; Saracevic, 1975; Swanson, 1986) called objective (system-based)
and subjective (human (user)-based) relevance respectively. The objective relevance can be viewed as
a topicality measure, i.e. a direct match of the topic of the retrieved document and the one defined by
the query. Several studies on the human relevance show that many other criteria are involved in the
evaluation of the IR process output (Barry, 1998; Park, 1993; Vakkari & Hakala, 2000). In particular the
subjective relevance refers to the intellectual interpretations carried out by users and it is related to the
concepts of aboutness and appropriateness of retrieved information. According to Saracevic (1996) five
types of relevance exist: an algorithmic relevance between the query and the set of retrieved information
objects; a topicality-like type, associated with the concept of aboutness; cognitive relevance, related to
the user information need; situational relevance, depending on the task interpretation; and motivational
and affective relevance, which is goal-oriented. Furthermore, we can say that relevance has two main
features defined at a general level: multidimensional relevance, which refers to how relevance can be
perceived and assessed differently by different users; dynamic relevance, which instead refers to how
this perception can change over time for the same user. These features have great impact on information
retrieval systems which generally have not a user model and are not adaptive to individual users.
It is generally acknowledged that some techniques can help the user in information retrieval tasks
with more awareness, such as relevance feedback (RF). Relevance feedback is a means of providing
additional information to an information retrieval system by using a set of results provided by a classical
system by means of a query (Salton & Buckley, 1990). In the RF context, the user feeds some judgment
back to the system to improve the initial search results. The system can use this information to retrieve
other documents similar to the relevant ones or ranks the documents on the basis of user clues. In this
chapter we describe a system which uses the second approach. A user may provide the system with
relevance information in several ways. He may perform an explicit feedback task, directly selecting
documents from list results, or an implicit feedback task, where the system tries to estimate the user
interests using the relevant documents in the collection. Another well known technique is the blind (or
pseudo) relevance feedback where the system chooses the top-ranked documents as the relevant ones.
LITer ATure oVer VIeW
Relevance feedback techniques have been investigated for more then 30 years (Spink & Losee, 1996)
and several papers show that they are effective for improving retrieval performance (Harman, 1992;
Rocchio, 1971). From a general point of view RF techniques refer to the measure of relevance. In this
context an end-user bases his judgment on the expected contribution of the analyzed document to his
task. Resnick et al., (1994) presents GroupLens, a collaborative filter-based system which ranks the docu-
ments on the basis of numeric ratings explicitly assigned by the user. The basic idea is that people who
agreed with the evaluation of past articles are likely to agree again in the future. Moreover the system
tries to predict user's agreement using the ratings from similar users. SIFT (Yan & Garcia-Molina, 1995)
approach requires the user to explicitly submit his profile and update it using relevance feedback. The
SIFT engine uses profiles to filter documents and notifies them according to user-specified parameters.
AntWorld (Kantor et al., 2000) pursues the ant metaphor allowing Internet users to get information
about other users' quests. The users have to give a judgment about the visited pages. The judgment
is expressed using textual annotation and numeric value. The quests are stored in the system and the
Search WWH ::
Custom Search