Abstract:

Aiming at the term mismatch issues of the existing information retrieval systems, a novel query expansion algorithm of local feedback is proposed based on the fusion of negative association rules mining and feature terms extraction. Firstly, the feature terms from the topranked n retrieved local documents are extracted to construct the feature terms database, and the frequent itemsets and nonfrequent itemsets containing original query terms and non query terms synchronously are mined in the feature terms database. And then, the negative association rules, the antecedent of which is the original query terms, are mined in frequent itemsets and nonfrequent itemsets, and the consequents of the negative association rules are extracted as the negative association terms, and the correlation of each negative association term and the entire original query is calculated. Finally, the terms which are the same as the negative association terms are removed from the feature terms database according to the correlation and the rest of the terms of the feature terms database are combined with the original query for query expansion. The results of the experiment show that the proposed algorithm can effectively improve and enhance the information retrieval performance.

Key words: query expansion;local feedback;feature term;negative association rule