Genetic algorithm based dynamic parameter learning for text retrieval

In information retrieval (IR) systems, such as Okapi, there are always a variety of parameters to be set manually which are data-dependent and most sensitive to retrieval performance. Therefore, it will be ideal to deploy an automatic parameter learning mechanism. In this paper, we propose such a method based on the genetic algorithm. We apply our approach to the Okapi system. Experimental results on TREC2001 testing data indicate that our algorithm is effective to adjust system parameters and improve the retrieval performance significantly.