论文信息 - Relevance Feedback by Exploring the Different Feedback Sources and Collection Structure

Relevance Feedback by Exploring the Different Feedback Sources and Collection Structure

In HARD track of HARD 2005, we classify the 50 queries into 7 categories and make use of 3 kinds of feedback sources in various tasks. We find that the different kinds of queries perform differently in feedback tasks and the "CASE " and "EVENT" queries are more sensitive to the feedback source. We also explore the internal structure of corpus and try to estimate the distribution of relevant documents within sub-collections. The experiments show that this technology is partly effective and the main existing problem is how to predict the distribution more precisely.

Wei Zhang | Junlin Zhang | Le Sun | Yuanhua Lv

[1] Michael McGill,et al. Introduction to Modern Information Retrieval , 1983 .

[2] Terry Sullivan. Locating question difficulty through explorations in question space , 2001, JCDL '01.

[3] Ellen M. Voorhees,et al. Overview of the TREC 2004 Robust Retrieval Track , 2004 .

[4] C. J. van Rijsbergen,et al. Probabilistic models of information retrieval based on measuring the divergence from randomness , 2002, TOIS.

[5] ChengXiang Zhai,et al. Error analysis of difficult TREC topics , 2003, SIGIR '03.

[6] Claudio Carpineto,et al. Query Difficulty, Robustness, and Selective Application of Query Expansion , 2004, ECIR.

[7] Gerard Salton,et al. The SMART Retrieval System—Experiments in Automatic Document Processing , 1971 .