Incorporating Passage Feature Within Language Model Framework for Information Retrieval

  • Authors:

  • Ke Dang;Tiejun Zhao;Haoliang Qi;Dequan Zheng

  • Affiliations:

  • MOE-MS Key Laboratory of Natural Language Processing and Speech, Harbin Institute of Technology, Harbin, 150001, China;MOE-MS Key Laboratory of Natural Language Processing and Speech, Harbin Institute of Technology, Harbin, 150001, China;MOE-MS Key Laboratory of Natural Language Processing and Speech, Harbin Institute of Technology, Harbin, 150001, China;MOE-MS Key Laboratory of Natural Language Processing and Speech, Harbin Institute of Technology, Harbin, 150001, China

  • Venue:
  • CICLing '07 Proceedings of the 8th International Conference on Computational Linguistics and Intelligent Text Processing
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

Passage feature has been proved very useful in document retrieval. In this paper, we successfully incorporate the passage feature into language model framework by extending the Jelinek-Mercer smoothing. This scheme not only increases the precision of document language model but also can let the passage feature act well in the documents that are not very long. We compare our schemes with 4 baselines: the unigram language model and the passage language model with Jelinek-Mercer and Dirichlet smoothing. Experimental results on the TREC collections indicate that our method significantly outperforms the unigram language model and gets better performance than passage language model in collections whose documents are not very long.