A probabilistic method for inferring preferences from clicks

By Katja Hofmann, Shimon Whiteson, Maarten de Rijke. 2011

127 citations recorded by the source corpus.

Subfields: Computer Science Applications, Information Systems, 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 Computer Science Applications

1. Maarten de Rijke
2. Masrour Zoghi
3. Shimon Whiteson
4. Katja Hofmann
5. Ilya Markov
6. Róbert Istvan Busa-Fekete
7. Eyke Hüllermeier
8. Adil El Mesaoudi-Paul

In Information Systems

1. Maarten de Rijke
2. Masrour Zoghi
3. Shimon Whiteson
4. Ilya Markov
5. Katja Hofmann
6. Miroslav Dud'ik
7. Chang Li
8. Aleksandrs Slivkins

In Information Retrieval & Search

1. Masrour Zoghi
2. Maarten de Rijke
3. Katja Hofmann
4. Shimon Whiteson
5. Ilya Markov
6. Eyke Hüllermeier
7. Róbert Istvan Busa-Fekete
8. Adil El Mesaoudi-Paul

Citations in the corpus

• Contextual Dueling Bandits (2015, methodology): “As earlier noted, compared to sparring FPL, this bound gives a different trade-off between m and ε.”
• Click Models for Web Search (2015, background): “For instance, Hofmann et al. [2013] consider three different extreme cases for the attractiveness parameter assignment, which they call perfect, navigational and informational user models.”
• Relative confidence sampling for efficient on-line ranker evaluation (2014, methodology): “To compare a pair of rankers, we use probabilistic inter-leave (PI) [14], though any other interleaved comparison method could be used instead.”
• Reusing historical interaction data for faster online learning to rank for IR (2013, methodology): “Compared to the evaluation setting, where PI was shown to outperform BI and TD when applied to compare rankers over large amounts of data, PI performs worse than expected [9].”
• MergeRUCB: A Method for Large-Scale Online Ranker Evaluation (2015, methodology): “To compare a pair of rankers, we use probabilistic interleave (PI) [14], though any other interleaved comparison method could be used instead.”
• MergeDTS (2018, methodology): “[54], who use PI [22] to obtain a preference matrix for the MSLR dataset, and obtain a preference matrix for the Yahoo datasets.”
• Fidelity, Soundness, and Efficiency of Interleaved Comparison Methods (2013, methodology): “Interleaved comparison methods [Chapelle et al. 2012; Hofmann et al. 2011; Radlinski and Craswell 2010; Radlinski et al. 2008b], which compare rankers using naturally occurring user interactions such as clicks, are quickly gaining interest as a complement to TREC-style evaluations.”
• Preference-based Online Learning with Dueling Bandits: A Survey (2018, methodology): “This is done by the probabilistic multileave comparison method (Schuth et al., 2015), an extension of probabilistic interleave (Hofmann et al., 2011), which selects documents from a distribution, where the probability of being added correlates with the perceived relevance.”
• Copeland Dueling Bandits (2015, methodology): “To compare a pair of rankers, we use probabilistic interleave (PI) [34], a recently developed method for interleaved comparisons.”