Optimizing Guided Traversal for Fast Learned Sparse Retrieval

Abstract

Recent studies show that BM25-driven dynamic index skipping can greatly accelerate MaxScore-based document retrieval based on the learned sparse representation derived by DeepImpact. This paper investigates the effectiveness of such a traversal guidance strategy during top k retrieval when using other models such as SPLADE and uniCOIL, and finds that unconstrained BM25-driven skipping could have a visible relevance degradation when the BM25 model is not well aligned with a learned weight model or when retrieval depth k is small. This paper generalizes the previous work and optimizes the BM25 guided index traversal with a two-level pruning control scheme and model alignment for fast retrieval using a sparse representation. Although there can be a cost of increased latency, the proposed scheme is much faster than the original MaxScore method without BM25 guidance while retaining the relevance effectiveness. This paper analyzes the competitiveness of this two-level pruning scheme, and evaluates its tradeoff in ranking relevance and time efficiency when searching several test datasets.

This paper is published in WWW'23

References in corpus (8)

• Overview of the TREC 2020 deep learning track
• An Efficiency Study for SPLADE Models
• A Few Brief Notes on DeepImpact, COIL, and a Conceptual Framework for Information Retrieval Techniques
• SparTerm: Learning Term-based Sparse Representation for Fast Text Retrieval
• A Replication Study of Dense Passage Retriever
• Wacky Weights in Learned Sparse Representations and the Revenge of Score-at-a-Time Query Evaluation
• From Distillation to Hard Negative Sampling: Making Sparse Neural IR Models More Effective
• Dual Skipping Guidance for Document Retrieval with Learned Sparse Representations