“School of Cognitive”

Back to Papers Home
Back to Papers of School of Cognitive

Paper   IPM / Cognitive / 9578
School of Cognitive Sciences
  Title:   Evolving Weighting Functions For Query Expansion Based on Relevance Feedback
  Author(s): 
1.  Ali Borji
2.  M.Z. Jahromi
  Status:   In Proceedings
  Proceeding: Lecture Notes in Computer Science, APWEB
  Year:  2008
  Supported by:  IPM
  Abstract:
A new method for query expansion using genetic programming (GP) is proposed in this paper to enhance the retrieval performance of text information retrieval systems. Using a set of queries and retrieved relevant and non-relevant documents corresponding to each query, GP tries to evolve a criteria for selecting terms which when added to the original query improve the next retrieved set of documents. Two experiments are conducted to evaluate the proposed method over three standard datasets: Cranfield, Lisa and Medline. In first experiment a formula is evolved using GP over a training set and is then evaluated over a test query set of the same dataset. In the second experiment, evolved expansion formula over a dataset is evaluated over a different dataset. We compared our method against the base probabilistic method in literature. Results show a higher performance in comparison with original and probabilistically expanded method.

Download TeX format
back to top
scroll left or right