Fitting score distribution for blog opinion retrieval

ABSTRACT

Current blog opinion retrieval approaches cannot be applied if the topic relevance and opinion score distributions by rank are dissimilar. This problem severely limits the feasibility of these approaches. We propose to tackle this problem by fitting the distribution of opinion scores, which replaces the original topic relevance score distribution with the simulated one. Our proposed score distribution fitting method markedly enhances the feasibility of a state-of-the-art dictionary-based opinion retrieval approach. Evaluation on a standard TREC blog test collection shows significant improvements over high quality topic relevance baselines.

REFERENCES

• 1: G. Amati. Probabilistic models for information retrieval based on Divergence from Randomness. PhD thesis, University of Glasgow, 2003.
• 2: Ben He , Craig Macdonald , Jiyin He , Iadh Ounis, An effective statistical approach to blog post opinion retrieval, Proceeding of the 17th ACM conference on Information and knowledge management, October 26-30, 2008, Napa Valley, California, USA [doi>10.1145/1458082.1458223]
• 3: I. Ounis, C. Macdonald, and I. Soboroff. Overview of the TREC 2008 Blog Track. In Proceedings of TREC 2008.

INDEX TERMS

Primary Classification:
  H. Information Systems
  H.3 INFORMATION STORAGE AND RETRIEVAL
      H.3.3 Information Search and Retrieval

General Terms:
Experimentation, Performance

Keywords:
blog search, distribution fitting, opinion finding