Probabilistic Models in Information Retrieval

  • Citation-Key:
  • Title:
    Probabilistic Models in Information Retrieval
  • Author(s):
    N. Fuhr
  • Journal:
    The Computer Journal
  • Volume:
  • Number:
  • Page(s):
  • Year:


Information retrieval systems (IRS) are interactive information systems for vague queries and uncertain knowledge. The application range is not restricted to text retrieval, since the approach can be applied to other forms of knowledge as well. In this paper, a conceptual model for IRS is presented in which a distiction is made between the logic, the layout and the semantic view of database objects. Different methods for improving the semantic view are discussed. For the intrinsic uncertainty of IRS, a generalization of the proof-theoretic view on databases is presented; this approach is based on a probabilistic logic. Due to the uncertainty, IRS have to be used mostly in interactive mode. Here the user has to consider the output resulting from each dialogue step before he decides on the further proceeding. For this purpose, IR systems need a variety of interactive functions.
Subject descriptor(s):
retrieval models
survey, probabilistic retrieval

Fulltext as PS

Fulltext as PDF