The probability ranking principle in IR

By Stephen E. Robertson. 1997

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

Learning to rank for information retrieval (2009, unknown): “[11] F.”
Relevance-Based Language Models (2001, methodology): “Earlier work on probabilistic models of information retrieval [19, 18, 17, 22] took a conceptually different approach.”
Learning diverse rankings with multi-armed bandits (2008, background): “Although some researchers have transformed usage data into relevance judgments, or used it to generate features (e.g. Joachims, 2002; Radlinski & Joachims, 2005; Agichtein et al., 2006), we go one step further by directly optimizing a usage-based metric.”
Fairness of Exposure in Rankings (2018, background): “yu(d|q)leads to the ranking that maximizes the utility argmax r U(r|q)≡argsort d∈Du(d|q) for any function vthat decreases with rank. This is the insight behind the Probability Ranking Principle (PRP) [28]. 3.2 Probabilistic Rankings Rankings are combinatorial objects, such that naively searching the”
Beyond independent relevance: methods and evaluation metrics for subtopic retrieval (2003, background): “metric) for each document separately, and then ranking docu ments by probability of relevance [11].”
A Brief Survey of Text Mining (2005, methodology): “The currently predominant approaches based on this idea are the vector space model [SWY75], the probabilistic model [ Rob77 ] and the logical model [van86].”
Cross-lingual relevance models (2002, methodology): “Lavrenko and Croft [12] advocate using the Probability Ranking Principle [14], where documents are ranked by the probability ratio: P (D|R)/P (D|N).”
Novelty and diversity in information retrieval evaluation (2008, background): “The probability ranking principle (PRP) forms the bedrock of information retrieval research [22,24].”
Ad hoc classification of radiology reports. (1999, methodology): “Using a probabilistic methodology, Inquery ranks retrieved documents according to their likelihood of being relevant to the information need represented by a query.(2)”
The Turn - Integration of Information Seeking and Retrieval in Context (2005, background): “There are unrealistic assumptions regarding document independence (some may be relevant only if juxtaposed) and user saturation (repeated reproduction of very similar “relevant” information results in irrelevance in real-life situations) (Robertson 1977).”
A support vector method for optimizing average precision (2007, unknown): “Copyright 2007 ACM 978-1-59593-597-7/07/0007 ...$5.00. a query (e.g., [ 18 , 14]).”
Interpreting TF-IDF term weights as making relevance decisions (2008, methodology): “The previous basic ranking formula is consistent with the probability ranking principle [Robertson 1977] because it ranks documents by probability of relevance.”
Methods for Evaluating Interactive Information Retrieval Systems with Users (2009, background): “This is not to claim that there are no theories or theoretical constructs in IR and IIR — some examples include Robertson’s [218] probability ranking principle, Ingwersen’s [137] theory of polyrepresentation and Belkin’s [19] anomalous states of knowledge.”
Display Advertising with Real-Time Bidding (RTB) and Behavioural Targeting (2016, background): “A rich literature can be found from the probability ranking principle [Robertson, 1977], the RSJ and BM25 model [Jones et al., 2000], language models of IR [Ponte and Croft, 1998], to the latest development of learning to rank [Joachims, 2002, Liu, 2009], results diversity [Wang and Zhu, 2009,…”
Retrieval-Enhanced Machine Learning (2022, methodology): “Existing retrieval models are mostly designed based on the probability ranking principle (PRP) [49], in which documents are ranked based on their probability of relevance to the query.”