Okapi at TREC-3

By Stephen E. Robertson, S. Walker, S. Jones, Micheline Hancock-Beaulieu, M. Gatford. 1994

Subfields: 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 Information Retrieval & Search

Citations in the corpus

Distilling Knowledge from Reader to Retriever for Question Answering (2020, methodology): “This technique was later extended, leading to the BM25 weighting scheme which is still widely used today (Robertson et al., 1995).”
Fact, Fetch, and Reason: A Unified Evaluation of Retrieval-Augmented Generation (2024, background): “(2) BM25-Retrieved Prompt (n_docs): This approach augments the question with the top n_docs documents having the highest BM25 score (Robertson et al., 1995) retrieved from a Wikipedia data dump 2 .”
deltaBLEU: A Discriminative Metric for Generation Tasks with Intrinsically Diverse Targets (2015, methodology): “To this end, we employed an information retrieval algorithm with a bag-of-words BM25 similarity function (Robertson et al., 1995), as detailed in Sordoni et al.”
Multi-Stage Document Ranking with BERT (2019, methodology): “In our implementation, the query is treated as a “bag of words” for ranking documents from the corpus using a standard inverted index based on the BM25 scoring function (Robertson et al., 1994).”
Document Expansion by Query Prediction (2019, methodology): “However, due to the size of most corpora and the impracticality of applying inference over every document in response to a query, nearly all implementations today deploy neural networks as re-rankers over initial candidate sets retrieved using standard inverted indexes and a term-based ranking model”
Deep Learning for Biomedical Information Retrieval: Learning Textual Relevance from Click Logs (2017, background): “We believe the better performance of BM25 here is partly due to it being a strong factor in the relevance ranking from which these click logs were extracted, thus biasing the click data to some extent.”
Medical Question Answering for Clinical Decision Support (2016, unknown): “BM25 relied on the Okapi-BM25 [23] ( k 1 =1 .”
Toward Reproducible Baselines: The Open-Source IR Reproducibility Challenge (2016, unknown): “75 as recommended by Robertson [15].”
Title language model for information retrieval (2002, methodology): “The two baseline methods are the Okapi method[9] and the traditional language modeling approach.”
A Study of Information Retrieval Weighting Schemes for Sentiment Analysis (2010, background): “Robertson et al. (2004)) according to which, the probability of a document being relevant to a query term is typically sublinear in relation to the number of times a query term appears in the document.”
Inferring Query Performance Using Pre-retrieval Predictors (2004, methodology): “As one of the most well-established IR systems, Okapi uses BM25 to measure the term weight, where the idf factor w(1) is normalised as follows [ 12 ]:”
Span Selection Pre-training for Question Answering (2019, background): “The passage is relevant as determined by a BM25 (Robertson et al., 1995) (k1=1.”
PubMed related articles: a probabilistic topic-based model for content similarity (2007, unknown): “3.2 Comparison to Other Work A suitable point of comparison for this work is the Binary Independent Retrieval (BIR) model for probabilistic IR [ 5 ,6], which underlies bm25.”
Delaying Interaction Layers in Transformer-based Encoders for Efficient Open Domain Question Answering (2020, methodology): “ect¨ prelevant documents and only apply the costly reading comprehension model on them. Such a combination has proven itself in BertSerini (Yang et al. 2019a) where the widely known Lucene with BM25 (Robertson et al. 1995; Białecki et al. 2012) for the IR part was combined with Bert for the eQA part”