A cluster-based resampling method for pseudo-relevance feedback

摘要: Typical pseudo-relevance feedback methods assume the top-retrieved documents are relevant and use these pseudo-relevant to expand terms. The initial retrieval set can, however, contain a great deal of noise. In this paper, we present cluster-based resampling method select better based on relevance model. main idea is document clusters find dominant for set, repeatedly feed emphasize core topics query. Experimental results large-scale web TREC collections show significant improvements over For justification approach, examine density documents. A higher will result in greater accuracy, ultimately approaching true feedback. approach shows than baseline model all collections, resulting accuracy This indicates that proposed effective