Chapter Twelve - Diversity and novelty in biomedical information retrieval

Abstract

In this chapter, we review biomedical information retrieval techniques and focus on diversity and novelty boosting methods and their evaluation metrics. We introduce three diversity and novelty boosting approaches including maximal marginal relevance, probabilistic latent semantic analysis and relevance-novelty graphical model, and three diversity and novelty assessment measures including weighted subtopic precision, α-nDCG, and geNov. Experimental results on a set of state-of-the-art diversity and novelty evaluation metrics are also presented with regard to their sensitiveness to ranking qualities, their discriminative powers, and time efficiencies. We also conduct experiments with a larger dataset to reexamine these diversity and/or novelty metrics and present the results.