Interactive Information Retrieval

1
Interactive Information Retrieval
Vijaya Gambheera University at Albany
2
References
Ask for Information Retrieval Part I. Background
and Theory N.J. Belkin, R.N. Oddy and H.M.
Brooks, The journal of Documentation, Vol. 38,
No.2, 1982, 299-304 Comparing Interactive
Information Retrieval Systems across Sites The
TREC-6 Interactive Track Matrix Experiment E.
Lagergren, P. Over, SIGIR98, Melbourne,
Australia, ACM 1-58113-015-5 8/98, 164-172
3
Work by Belkin et al. has called into question
some traditional assumptions of IR concerning
The relationship of the request put to the
IR system to the information need underlying
the request The basis for text and request
representation in IR systems Retrieval
mechanisms suitable for IR
4

• The Ask hypothesis
• An information need arises from a
• recognized anomaly in the users state of
• knowledge concerning some topic or situation
• and that, in general, the user is unable to
• specify precisely what is needed to resolve that
• anomaly. Thus, for the purposes of IR, it is
• more suitable to attempt to describe that ASK,
• than to ask the user to specify her/his need as
• a request to the system.

5

• Best-Match Principle
• Given a representation of a request for
  information (e.g. Query), the best possible
  system
• response will be the text whose representation
  most closely matches it.
• Assumptions
• It is possible for the user to specify
  precisely the information that she/he requires.
• Information needs are functionally equivalent
  to document texts.

6

• Why Best-Match Principle is not always good?
• Information need is in fact not a need in itself,
  but rather a means toward satisfying some more
  basic need, typically, the resolution of a
  problem.
• A document is a representation of a coherent
  state of knowledge, while a query related to an
  information need will be a representation of an
  anomalous or inadequate or incoherent state of
  knowledge
• Associative retrieval and relevance feed back
  techniques are used to resolve the above problems

7

• - Hollnegels Cognitive Communication Model
  suggests that the user and system can be
  considered as partners in a dialogue.
• Accurate models and dialogue in general be best
  arrived at through a series of interactions,
  rather than just an initial statement and
  subsequent response.
• Ex Relevance feed back

8

• Structure of ASK IR System
• IR systems in general, consists of
• A mechanism for representing information need
• A text store
• A mechanism for representing and organizing
• texts
• A mechanism for retrieving texts
• A mechanism for evaluating the effectiveness
• of the retrieval

9
(No Transcript)
10

• Means of representing Information
• A network of concepts represented by words or by
  association data.
• This is based on the assumption that there is no
  real or true representation of
  knowledge/information, but rather many possible
  representations, each appropriate to particular
  problems.

11

• Retrieval Mechanism
• The basis of any evaluation must be the
  satisfaction of the user of the IR system with
  the systems performance in resolution of his/her
  ASK.
• -Oddy built THOMAS system which is truly
  interactive in that each retrieval depends upon
  the users evaluation of the results of the
  previous retrieval

12

• The outline of the proposed IR System
• Users problem statement
• Structural analysis of problem statement
• Choice of the retrieval strategy according to
  type of ASK.
• Abstract presented to the user simultaneously
  with explanation of why text was chosen.
• Structured dialogue between system and user for
  system to infer users evaluation of
• a) Method of choice
• b) Suitability of document to problem

13

• The system would work as follows
• The user discusses his/her information problem in
  an unstructured statement.
• The problem statement is converted to a
  structural representation of the users ASK by
  the text analysis program.
• One of several available retrieval mechanisms is
  chosen to interrogate the database.
• The abstract is printed out for the reader to
  read.

14
5. The system then initiates a structured
dialogue with the user inferring from the
response the users attitude toward a) The
method of choice b) The suitability of the
text to the problem c) Whether his/her
information need has changed. 6. a) The system
changes retrieval mechanism if necessary, and/or
b) The system modifies the problem structure
if necessary, or c) The system stops if the
user is satisfied. 7. The system returns to Step
2 or 3.
15
Participants -No searcher could have used
either the control system or the experimental
system. Apparatus -Each site runs their
experimental system(s) Computing resources
-Adequate to run both the control and
experimental systems and collect the data
required.
16
Topics Ferry sinkings International art
crime New hydroelectric projects Wildlife
extinctions Hubble telescope achievements
Alzheimers drug treatment
17
General Findings The results confirm the
importance of applying good experimental design
principles to extract maximal information from
interactive IR experiments. The system should
allow for more searcher-topic and searcher-system
interactions to improve the performance The
lack of strong searcher affected almost half of
the sites
18
Thank You!!!