Query Modification Based on Relevance Feedback
Wafaa ALhamed · 2019
Query modification based on the pseudo relevance feedback(PRF) is an effective way for improving the performance of retrieval system. It is promising approach that used to solve the problems of ambiguous language and synonyms depending on the feedback. In this paper each of expansion and reduction approaches are used for query modification. A hybrid of distribution and association methods are used to select significant terms and expand the query. In this paper, Kullback-Leibler Divergence (KLD) and CHI squared were used as distribution methods that depend on the terms distribution at the top relevant documents and distribute them in the entire corpus. In addition, proximity term frequency method was used to show the semantic relationship between each candidate term with query terms. Different sets of candidate terms are produced from these methods will be combined by Borda aggregation method to get the strength of each individual method. The second direction is removing insignificant terms from the original query and terms' pool to prevent them from being used for expansion query.