Dynamic Superblock Pruning for Fast Learned Sparse Retrieval
About
This paper proposes superblock pruning (SP) during top-k online document retrieval for learned sparse representations. SP structures the sparse index as a set of superblocks on a sequence of document blocks and conducts a superblock-level selection to decide if some superblocks can be pruned before visiting their child blocks. SP generalizes the previous flat block or cluster-based pruning, allowing the early detection of groups of documents that cannot or are less likely to appear in the final top-k list. SP can accelerate sparse retrieval in a rank-safe or approximate manner under a high-relevance competitiveness constraint. Our experiments show that the proposed scheme significantly outperforms state-of-the-art baselines on MS MARCO passages on a single-threaded CPU.