A dynamic bayesian network click model for web search ranking

By Olivier Chapelle, Ya Zhang. 2009

597 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Information Systems

How researchers see this paper

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

In Computer Vision and Pattern Recognition

In Information Systems

Citations in the corpus

Counterfactual Estimation and Optimization of Click Metrics for Search Engines (2014, background): “Prediction errors of state-of-the-art user click models ( e.g. , [9, 14] and their variants) are likely much larger than usual improvements of click-based metrics in nowadays commercial search engines that are already highly optimized.”
Off-policy evaluation for slate recommendation (2016, background): “Another line of work constructs parametric models of slate quality [8, 16, 14] (see also Sec.”
A General Framework for Counterfactual Learning-to-Rank (2018, background): “Other click models ([3, 7, 9], also see [8]) train to maximize loglikelihood of observed clicks, where relevance is modeled as a latent variable that is inferred over multiple instances of the same query.”
Big Data or Right Data? (2013, background): “One of the most well-known examples is click data in search engines, where the data is biased by the ranking and the user interface (see for example [10, 7]).”
Click chain model in web search (2009, methodology): “A most recent related work is a DBN click model proposed by Chapelle and Zhang [3], which appears as another paper in this proceeding.”
Time-Aware Click Model (2016, methodology): “The dynamic Bayesian network model (DBN) [Chapelle and Zhang 2009] is the first model to consider presentation bias due to a snippet (rather than ranking position).”
Human Curation and Convnets: Powering Item-to-Item Recommendations on Pinterest (2015, methodology): “The relevance triplets we use in training are generated through user clicks and impression logs, normalized by position and device to account for position bias, using a clicks over expected clicks [3] (COEC) model.”
Toward Predicting the Outcome of an A/B Experiment for Search Relevance (2015, background): “Much progress has been made in the past that led to reﬁned models that deepen our understanding of user engagement ( e.g. , [8, 10, 11]).”
Learning to Rank with Selection Bias in Personal Search (2016, methodology): “In order to estimate these marginals, the click models often assume access to large quantities of click data for each document x given the query [8, 11, 15].”
Counterfactual Estimation and Optimization of Click Metrics in Search Engines: A Case Study (2015, background): “Prediction errors of state-of-the-art user click models ( e.g. , [7, 12] and their variants) are likely much larger than usual improvements of click-based metrics in nowadays commercial search engines that are already highly optimized.”
Bias on the web (2018, background): “11,12 Other-wise, the popular pages become even more popular.”
Transfer Learning for Behavior Ranking (2017, methodology): “Notice that the seemly very similar click models in information retrieval such as the cascade model [Craswell et al. 2008], position model [Chapelle and Zhang 2009], and general click model [Allen Zhu et al. 2010] actually differ: (i) they do not answer the above ﬁrst question, because they require…”
Bias and Debias in Recommender System: A Survey and Future Directions (2020, background): “Training Heuristic [69], [70], [71], [72], [73], [74], [75], [76] Sampling [30], [43], [74], [77], [78], [79] Exposure-based model [7], [8], [30], [31], [80] Others [5], [6], [81], [82], [83], [84], [85] Position Bias Click models [86], [87], [88], [89], [90], [91], [92] Propensity Score [5], [93], ”
The effect of ad rank on the performance of keyword advertising campaigns (2013, background): “Noticing such rank bias, researchers have proposed several models aiming to better explain and eliminate such serial position effect (Craswell, Zoeter, Taylor, & Ramsey, 2008; Chapelle & Zhang, 2009).”
Expected reciprocal rank for graded relevance (2009, background): “The original cascade model of [12] has later been extended in [8] to include an abandonment probability: if the user is not satisﬁed at a given position, he will examine the next url with probability γ , but has a probability 1 − γ of aban-doning.”
Recommending what video to watch next: a multitask ranking system (2019, methodology): “One method to remove position bias is inspired by [8], where Chapelle et al. evaluated a CTR model using P ( relevance | pos = 1 ) , under the assumption of no position bias efect for evaluation at position 1.”
Unbiased Learning-to-Rank with Biased Feedback (2016, background): “Other click models ([5, 3, 1], also see [4]) have been proposed, and are trained to maximize log-likelihood of observed clicks.”