A New Hybrid Document Clustering for PRF-Based Automatic Query Expansion Approach for Effective IR

Related Work

A few researchers have proposed document clustering to enhance the performance of IR. Lee et al. (2000) used document clustering approach for PRF based document selection. They considered a document pseudo-relevant document only when it has high likeness with other documents and low similarity or no similarity with neighbor documents. This approach cannot be used for large diversified documents. For example, if any long document contains unclear and general terms (Fall et al., 2003) then there is a higher possibility of choosing that document for PRF against most of the queries. Moreover, if the documents are very short and contain many synonyms then the size of the cluster will be small and these documents will have high possibility to be chosen for PRF. Therefore, existing clustering approaches do not provide exact distribution of documents for PRF.