The Probabilistic Relevance Framework: BM25 and Beyond

By Stephen E. Robertson, Hugo Zaragoza. 2009

5,679 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Statistics and Probability, Information Retrieval & Search

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Statistics and Probability

In Information Retrieval & Search

Citations in the corpus

Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks (2020, methodology): “We also compare RAG’s dense embedding-based retrieval mechanism to a word overlap-based BM25 retriever [47].”
Improving language models by retrieving from trillions of tokens (2021, methodology): “Historically, information retrieval for text relies on inverted index matching such as TF-IDF and BM25 (Robertson and Zaragoza, 2009).”
Augmented Language Models: a Survey (2023, background): “Sparse retrievers work with sparse bag-of-words representations of the documents and the queries (Robertson and Zaragoza, 2009).”
Progressively Pretrained Dense Corpus Index for Open-Domain Question Answering (2020, methodology): “This kind of reranker can usually provide more accurate paragraph scores than the TF-IDF or BM25 based retriever while ranking the answer candidates.”
Re2G: Retrieve, Rerank, Generate (2022, background): “For example, the scores returned by BM25 [Robertson and Zaragoza, 2009] are not comparable to the inner products from DPR.”
Improving Wikipedia verifiability with AI (2022, background): “…et al., 2019; Mao et al., 2020) to translate the citation context into query text , and then matches the resulting query—a sparse bag-of-words vector—on a BM25 index (Robertson et al., 1995; Baeza-Yates et al., 1999; Manning et al., 2008; Robertson & Zaragoza, 2009; Lin et al., 2021) of Sphere.”
Autoregressive Search Engines: Generating Substrings as Document Identifiers (2022, methodology): “Unless otherwise speciﬁed, we use pyserini to compute the BM25 baseline.”
The Web Is Your Oyster - Knowledge-Intensive NLP against a Very Large Web Corpus (2021, methodology): “We consider two baseline retrieval models: BM25 (Robertson, 2009) and DPR (Karpukhin et al.”
Unsupervised Dense Information Retrieval with Contrastive Learning (2021, methodology): “Traditionally, retrieval systems, or retrievers, leverage lexical similarities to match queries and documents, using, for instance, TF-IDF or BM25 weighting (Robertson & Zaragoza, 2009).”
Learning to Retrieve Passages without Supervision (2021, methodology): “A Baselines: Further Details BM25 (Robertson and Zaragoza, 2009) A sparse bag-of-words model that extends TF-IDF (i.e. reward rare terms that appear in both q and p ) by accounting for document length and term frequency saturation.”
Adaptive Information Seeking for Open-Domain Question Answering (2021, background): “Implementation Details For sparse retrieval, we index all passages in the corpus with Elasticsearch and implement BM25 following Qi et al. (2019) 3 .”
The ROOTS Search Tool: Data Transparency for LLMs (2023, background): “We build 13 sparse, BM25 (Robertson, 2009) indices: one per language group for the”