Lune

EMNLP Findings2022

Accelerating Learned Sparse Indexes Via Term Impact Decomposition

Joel Mackenzie, Antonio Mallia, Alistair Moffat, Matthias Petri

Abstract

Novel inverted index-based learned sparse ranking models provide more effective, but less efficient, retrieval performance compared to traditional ranking models like BM25. In this paper, we introduce a technique we call postings clipping to improve the query efficiency of learned representations. Our technique amplifies the benefit of dynamic pruning query processing techniques by accounting for changes in term importance distributions of learned ranking models. The new clipping mechanism accelerates top-k retrieval by up to 9.6× without any loss in effectiveness.

Ask Lune

Know this paper in minutes, not an afternoon.

Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.

2 top-tier papers have cited it since. Know what they found before your reviewers do.

Questions to start from

Lune can draw on

the full paper · the top-tier papers that cite it · related work

Get the answer

Free to start. No credit card required.

Lineage

  • TILDE: Term Independent Likelihood moDEl for Passage Re-ranking cites ColBERT: Efficient and Effective Passage Search via Contextualized Late Interaction over BERT
  • Balancing the Blend: An Experimental Analysis of Trade-offs in Hybrid Search cites Dense Passage Retrieval for Open-Domain Question Answering
  • Balancing the Blend: An Experimental Analysis of Trade-offs in Hybrid Search cites ColBERT: Efficient and Effective Passage Search via Contextualized Late Interaction over BERT
  • SDR: Efficient Neural Re-ranking using Succinct Document Representation cites Dense Passage Retrieval for Open-Domain Question Answering
  • SDR: Efficient Neural Re-ranking using Succinct Document Representation cites ColBERT: Efficient and Effective Passage Search via Contextualized Late Interaction over BERT
  • Dense Passage Retrieval for Open-Domain Question Answering cites ColBERT: Efficient and Effective Passage Search via Contextualized Late Interaction over BERT
Builds onCited byLeft to right by year; the more cited, the higher.

Cited by top-tier papers2

  1. Balancing the Blend: An Experimental Analysis of Trade-offs in Hybrid SearchMengzhao Wang, Boyu Tan, Yunjun Gao, Hai Jin et al.VLDB 2026 · 8 citations
  2. Efficient Sparse Retrieval with Lightweight Superblock PruningParker Carlson, Wentai Xie, Rohil Shah, Tao YangSIGIR 2026 · 2 citations

Builds on4

  1. ColBERT: Efficient and Effective Passage Search via Contextualized Late Interaction over BERTOmar Khattab, Matei ZahariaSIGIR 2020 · 1,246 citations
  2. Dense Passage Retrieval for Open-Domain Question AnsweringVladimir Karpukhin, Barlas Oguz, Sewon Min, Patrick Lewis et al.EMNLP 2020 · 142 citations
  3. TILDE: Term Independent Likelihood moDEl for Passage Re-rankingShengyao Zhuang, Guido ZucconSIGIR 2021 · 86 citations
  4. SDR: Efficient Neural Re-ranking using Succinct Document RepresentationNachshon Cohen, Amit Portnoy, Besnik Fetahu, Amir IngberACL 2022

Related papers

  1. Optimizing Guided Traversal for Fast Learned Sparse RetrievalYifan Qiao, Yingrui Yang, Haixin Lin, Tao YangWWW 2023 · 19 citations
  2. BLADE: Combining Vocabulary Pruning and Intermediate Pretraining for Scaleable Neural CLIRSuraj Nair, Eugene Yang, Dawn J. Lawrie, James Mayfield et al.SIGIR 2023 · 4 citations
  3. Threshold-driven Pruning with Segmented Maximum Term Weights for Approximate Cluster-based Sparse RetrievalYifan Qiao, Parker Carlson, Shanxiu He, Yingrui Yang et al.EMNLP 2024 · 7 citations
  4. Efficient Neural Ranking using Forward IndexesJurek Leonhardt, Koustav Rudra, Megha Khosla, Abhijit Anand et al.WWW 2022 · 16 citations
  5. Finding the Best of Both Worlds: Faster and More Robust Top-k Document RetrievalOmar Khattab, Mohammad Hammoud, Tamer ElsayedSIGIR 2020 · 12 citations
Dusk over the sea between two cliffs drawn in fine vertical lines