Efficient Inverted Indexes for Approximate Retrieval over Learned Sparse Representations

This article describes a novel retrieval algorithm developed for collections of sparse vectors. This work, which is published in the proceedings of SIGIR'24, is the culmination of a joint collaboration with my brilliant colleagues Rossano Venturini (of the University of Pisa), Cosimo Rulli and Franco Maria Nardini (both of the Italian National Research Council, ISTI-CNR). Our proposed algorithm marks a seismic shift in retrieval over learnt sparse embeddings of text collections in accuracy and speed. We are all equally proud of this paper!

Foundations of Vector Retrieval

A monograph devoted to the theoretical foundations of nearest neighbor search, with a discussion of key data structures and algorithms from this vast literature.