Feedback Ranking Method in Topic-Based Retrieval

Abstract:

Ranking has an extensive application in analyzing public opinions of social network (SN), such as searching the most hot topic or the most relevant articles that the user concerning. In these scenarios, due to the different requirements of users, there is need to rank the object set from different aspects and to re-rank the object set by integrating these different results to acquire a synthesize rank result.In this paper, we proposed a novel Feedback Ranking method, which lets two basic rankers learn from each other during the mutual process by providing each one's result as feedback to the other so as to boost the ranking performance. During the mutual ranking refinement process, we utilize iSRCC---an improvement on Spearman Rank Correlation to calculate the weight of each basic rankers dynamically. We apply this method into the article ranking problem on topic-query retrieval and evaluate its effectiveness on the TAC09 data set. Overall evaluation results are promising.

• Periodical: Applied Mechanics and Materials (Volume 339)
• Pages: 269-274
• DOI: https://doi.org/10.4028/www.scientific.net/AMM.339.269
• Online since: July 2013
• Authors: Fei Li, Jia Jia Huang, Min Peng, Rui Cai
• Keywords: Feedback Ranking, F-Rank, Mutual Learning, Ranking

References

[1] Furu Wei, Wenjie Li and Yanxiang He, "Co-Feedback Ranking for Query-Focused Summarization", Proceedings of the ACL-IJCNLP 2009 Conference Short Papers, pages 117–120al, 2011.

DOI: 10.3115/1667583.1667620

[2] Furu Wei , Wenjie Li , Wei Wang and Yanxiang He, ''iRANK: An Interactive Ranking Framework and It Application in Query-Focused Summarization", CIKM'09, November 2–6,2009, Hong Kong, China .

DOI: 10.1145/1645953.1646171

[3] Paul Over, Hoa Dang and Donna Harman. "2007 DUC in Context", Information Processing and Management, 43(6):1506-1520.

DOI: 10.1016/j.ipm.2007.01.019

[4] A. Blum and T. Mitchell., "Combining Labeled and Unlabeled Data with Co-Training.", CoLT, pp.92-100.We,1998.

[5] J. Pickens and G. Colovchinsky, "Ranked Feature Fusion Models for Ad Hoc Retrieval", CIKM, pp.893-900, 2008.

[6] MEI paper on Spearmans rank correlation coefficient, (2007)

[7] HANDBOOK OF BIOLOGICAL STATISTICS ,http://udel.edu/~mcdonald/statspearman.html

[8] TAC2009 Update Summarization Task, http://www.nist.gov/tac/2009/Summarization

[9] Michael Collins and Yoram Singer, "Unsupervised models for named entity classification", In Proceedings of the 1999 SIGDAT Conference on Empirical Methods in Natural Language Processing and Very Large,(1999)

[10] Deerwester, S., et al,'' Improving Information Retrieval with Latent Semantic Indexing'', Proceedings of the 51st Annual Meeting of the American Society for Information Science 25, 1988, p.36–40.

[11] Akiko, Aizawa, "An information theoretic perspective of tf-idf measures", Information Processing & Management , January 2003,Pages45-65

DOI: 10.1016/s0306-4573(02)00021-3