A deep-dive into information retrieval systems: statistical perspective

Abstract

Evaluation of user queries in order to provide highly-accurate search results within a time limited window has been the pillar-stone formula for designing any information retrieval system. In order to perform this task, various retrieval models like Boolean-model, vector-space model, language-processing model, etc. have been proposed by researchers over the years. The major challenge for any information retrieval system designer is to select the best model suited for the application under development. Selection of this model requires an algorithm which can fulfill the constraints of query processing delay, retrieval accuracy, and the desired level of precision and recall values. In order to assist such researchers, in this paper we have analyzed and evaluated the recent state-of-the-art retrieval systems, and compared them in terms of general precision, recall, accuracy and delay metrics. Based on this review, researchers can select algorithms which are most suited for their system, and also improve on them based on the recommendations which are mentioned about these algorithms in this text.