A Feedback-Based Self-Organizing Query Structure Optimization Algorithm

Abstract

Query is one of the most important factors that can directly influence the results of information retrieval (IR). However, the query is defined by the user and thus inevitably has the following two problems: (1) the user often cannot exactly represent their search intention via query terms; (2) the user cannot effectively select the weight of each query term based on its importance toward the query's meaning. The above two problems cause the two types of uncertainty of a query. In this paper, we define them as the uncertainty of the query structure and the uncertainty of the query parameter, respectively. To eliminate the above two types of uncertainty and solve the above two problems, this paper proposes a new algorithm which includes two parts: (1) a self-organizing query structure loop which expands the initial query by adding only one term within each loop based on feedback technology until it meets the terminating condition of expansion defined by the author, and (2) an optimization algorithm based on a genetic algorithm (GA) that optimizes the weights of the expanded query vector within each loop. This algorithm provides a method of finding the optimal number of query expansion terms and improving the precision and recall of the search results. The experiment results show the effectiveness of the proposed algorithm.