This chapter covers advanced information retrieval (IR) concepts and provides the algorithmic solutions to the problems in conventional IR systems. Feature selectors are algorithms applied to the data before the data reach the document ranking program. The method introduced differs from the existing methodologies, especially in that it uses an efficient feature selector 0/1 knapsack-based heuristic with generalization and relevance feedback approach to determine the usefulness of features and evaluates its effectiveness with three common approaches such as Markov random field (MRF) model, correlation coefficient, and count difference. Correlation-based feature selection uses a search algorithm along with a function to evaluate the merit of feature subsets. Term discrimination tries to measure the ability of a feature for distinguishing one document from the others in a collection. A new feature selection method is called count difference (CD), which is based on the difference between the relative document frequencies of a feature for both relevant and irrelevant classes.