32,772 Hits in 9.6 sec

An experimental comparison of click position-bias models

Nick Craswell, Onno Zoeter, Michael Taylor, Bill Ramsey
2008 Proceedings of the international conference on Web search and web data mining - WSDM '08  
Search engine click logs provide an invaluable source of relevance information, but this information is biased.  ...  This paper focuses on explaining that bias, modelling how probability of click depends on position. We propose four simple hypotheses about how position bias might arise.  ...  EXPLAINING POSITION BIAS We present several hypotheses for how users view a search results list, and each of these leads to a model of position bias.  ... 
doi:10.1145/1341531.1341545 dblp:conf/wsdm/CraswellZTR08 fatcat:c4uxkeviovfnljyvcxonmbawcy

Position Bias Estimation for Unbiased Learning to Rank in Personal Search

Xuanhui Wang, Nadav Golbandi, Michael Bendersky, Donald Metzler, Marc Najork
2018 Proceedings of the Eleventh ACM International Conference on Web Search and Data Mining - WSDM '18  
We propose a regressionbased Expectation-Maximization (EM) algorithm that is based on a position bias click model and that can handle highly sparse clicks in personal search.  ...  Existing approaches use search result randomization over a small percentage of production traffic to estimate the position bias.  ...  ACKNOWLEDGMENTS We thank Olivier Chapelle for his discussions on tradeoff among different position bias estimation methods and Roberto Bayardo for his suggestions on improving the paper presentation.  ... 
doi:10.1145/3159652.3159732 dblp:conf/wsdm/WangGBMN18 fatcat:shszbvbxpbeydit6ik5f5isn7q


Xinyi Dai, Jiawei Hou, Qing Liu, Yunjia Xi, Ruiming Tang, Weinan Zhang, Xiuqiang He, Jun Wang, Yong Yu
2020 Proceedings of the 29th ACM International Conference on Information & Knowledge Management  
With a position-aware deep click-through rate prediction model, we address the attention bias considering both query-level and item-level features.  ...  Due to the item-specific attention bias modeling, the optimization for expected utility corresponds to a maximum weight matching on the item-position bipartite graph.  ...  ACKNOWLEDGEMENT The corresponding author Weinan Zhang thanks the support of NSFC (61702327, 61772333, 61632017) . The work is also sponsored by Huawei Innovation Research Program.  ... 
doi:10.1145/3340531.3412756 dblp:conf/cikm/DaiHLXT0H0020 fatcat:ejnob5uzdfakro6zl4xlmzzpuu

Learning to Rank with Selection Bias in Personal Search

Xuanhui Wang, Michael Bendersky, Donald Metzler, Marc Najork
2016 Proceedings of the 39th International ACM SIGIR conference on Research and Development in Information Retrieval - SIGIR '16  
Though a large amount of click data can be easily collected by search engines, various biases make it difficult to fully leverage this type of data.  ...  These click models typically require a large quantity of clicks for each individual pair and this makes them difficult to apply in systems where click data is highly sparse due to personalized corpora  ...  Experimental Results In this section, we evaluate the position bias prediction models in a couple of different settings.  ... 
doi:10.1145/2911451.2911537 dblp:conf/sigir/WangBMN16 fatcat:lgtamh6u2jfebex5tylo7m7mde

Eliminating Search Intent Bias in Learning to Rank [article]

Yingcheng Sun and Richard Kolacinski and Kenneth Loparo
2020 arXiv   pre-print
In order to measure the effects of biases, many click models have been proposed in the literature.  ...  Experimental results demonstrate that after adopting the search intent hypothesis, click models can better interpret user clicks and substantially improve retrieval performance.  ...  In the experimental section, we introduce the datasets, comparison methods and evaluation metrics, as well as the experimental results.  ... 
arXiv:2002.03203v2 fatcat:mgh5j75ysnc2dhhvhavz4i3t44

An F-shape Click Model for Information Retrieval on Multi-block Mobile Pages [article]

Jianghao Lin, Lingyue Fu, Weiwen Liu, Ruiming Tang, Weinan Zhang, Rui Zhang, Yong Yu
2022 arXiv   pre-print
These findings lead to the design of a novel F-shape Click Model (FSCM), which serves as a general solution to multi-block mobile pages.  ...  Most click models focus on user behaviors towards a single list.  ...  The performance of a click model heavily depends on appropriate assumptions on users' behavior patterns to tackle different biases in click logs (e.g., position bias [10] , appearance bias [36] ).  ... 
arXiv:2206.08604v2 fatcat:auwsjnhkuzen5h4dmzxhsojtwu

Beyond ten blue links

Danqi Chen, Weizhu Chen, Haixun Wang, Zheng Chen, Qiang Yang
2012 Proceedings of the fifth ACM international conference on Web search and data mining - WSDM '12  
Click model has been positioned as an effective approach to interpret user click behavior in search engines.  ...  With these biases and an effective model to correct them, FCM is more accurate in characterizing user click behavior in federated search.  ...  Besides the position bias in existing click models, we focus on illustrating the other biases introduced by the federated search and designing a new model to correct these biases to infer a more accurate  ... 
doi:10.1145/2124295.2124351 dblp:conf/wsdm/ChenCWCY12 fatcat:5dwlw47tdjc3vgb4ca5eff4kmq

Efficient multiple-click models in web search

Fan Guo, Chao Liu, Yi Min Wang
2009 Proceedings of the Second ACM International Conference on Web Search and Data Mining - WSDM '09  
Previous eye-tracking experiments and studies on explaining position-bias of user clicks provide a spectrum of hypotheses and models on how an average user examines and possibly clicks web documents returned  ...  Extensive experimental studies demonstrate the gain of modeling multiple clicks and their dependencies.  ...  An experimental comparison of click position-bias models. In WSDM '08: Proceedings of the first ACM international conference on Web search and data mining, pages 87-94, 2008. [7] G. E. Dupret, V.  ... 
doi:10.1145/1498759.1498818 dblp:conf/wsdm/GuoLW09 fatcat:sdafyi2jmvet3mxnln2sivhphi

Unbiased Top-k Learning to Rank with Causal Likelihood Decomposition [article]

Haiyuan Zhao, Jun Xu, Xiao Zhang, Guohao Cai, Zhenhua Dong, Ji-Rong Wen
2022 arXiv   pre-print
By decomposing the log-likelihood of the biased data as an unbiased term that only related to relevance, plus other terms related to biases, CLD successfully detaches the relevance from position bias and  ...  Empirical studies also showed that CLD is robust to the variation of bias severity and the click noise.  ...  Causal view of the position bias Next, based on a causal graph, we will illustrate why ranking models are biased if they are trained with user clicks directly, and how to realize unbiased learning for  ... 
arXiv:2204.00815v1 fatcat:vfdx4xdvtfet7cg2ujfrihtg7a

Incorporating vertical results into search click models

Chao Wang, Yiqun Liu, Min Zhang, Shaoping Ma, Meihong Zheng, Jing Qian, Kuo Zhang
2013 Proceedings of the 36th international ACM SIGIR conference on Research and development in information retrieval - SIGIR '13  
As an effective approach to interpret users' click-through behavior as feedback information, most click models were designed to reduce the position bias and improve ranking performance of ordinary search  ...  Based on these findings, a novel click model considering these biases besides position bias was constructed to describe interaction with SERPs containing verticals.  ...  It indicates that VCM better models user behavior on these vertical results by incorporating more biases besides position bias. 13 and Table 5 present log-likelihood comparison results for VCM and UBM  ... 
doi:10.1145/2484028.2484036 dblp:conf/sigir/WangLZMZQZ13 fatcat:haalq3557ngyjdwoxo4vvvdyoy

Partial rewarding during clicker training does not improve naïve dogs' learning speed and induces a pessimistic-like affective state

Giulia Cimarelli, Julia Schoesswender, Roberta Vitiello, Ludwig Huber, Zsófia Virányi
2020 Animal Cognition  
Generally, emotional reactivity was positively associated with a more pessimistic bias.  ...  We clicker-trained two groups of dogs: one group received food after every click while the other group received food only 60% of the time.  ...  However, small methodological differences within the use of positive reinforcement-based methods could have a diverse impact on an animal's affective state (with a potentially smaller amplitude in comparison  ... 
doi:10.1007/s10071-020-01425-9 pmid:32897444 fatcat:hw2btbdmjvf6pom4nps4y5wk5a

Optimizing Ranking Models in an Online Setting [article]

Harrie Oosterhuis, Maarten de Rijke
2019 arXiv   pre-print
To date the only comparisons of the two approaches are limited to simulations with cascading click models and low levels of noise.  ...  Second, we estimate an upper and lower bound on the performance of methods by simulating both ideal user behavior and extremely difficult behavior, i.e., almost-random non-cascading user models.  ...  All content represents the opinion of the authors, which is not necessarily shared or endorsed by their respective employers and/or sponsors.  ... 
arXiv:1901.10262v1 fatcat:rztik3atnfd6hhmyvtkccfinmm

Evolution of Popularity Bias: Empirical Study and Debiasing [article]

Ziwei Zhu, Yun He, Xing Zhao, James Caverlee
2022 arXiv   pre-print
Concretely, we conduct an empirical study by simulation experiments to analyze popularity bias in the dynamic scenario and propose a dynamic debiasing strategy and a novel False Positive Correction method  ...  These works fail to take account of the dynamic nature of real-world recommendation process, leaving several important research questions unanswered: how does the popularity bias evolve in a dynamic scenario  ...  We can use the prediction 𝑟 from a biased model, such as an MF, as 𝜃 𝑢,𝑖 , FPC is still vulnerable to the model bias.  ... 
arXiv:2207.03372v2 fatcat:uoyg5jtrhvahlempuqtyhbedkm

Unbiased LambdaMART: An Unbiased Pairwise Learning-to-Rank Algorithm [article]

Ziniu Hu, Yang Wang, Qu Peng, Hang Li
2019 arXiv   pre-print
Most of the algorithms, based on the inverse propensity weighting (IPW) principle, first estimate the click bias at each position, and then train an unbiased ranker with the estimated biases using a learning-to-rank  ...  In this paper, we propose a novel algorithm, which can jointly estimate the biases at click positions and the biases at unclick positions, and learn an unbiased ranker.  ...  We also consider developing a more general debiasing method that can deal with not only position bias but also other types of bias such as presentation bias.  ... 
arXiv:1809.05818v2 fatcat:m5j535miavbino54ybgdhrbqfq

Disentangling the Effects of Social Signals [article]

Tad Hogg, Kristina Lerman
2016 arXiv   pre-print
Using Amazon Mechanical Turk, we experimentally measure the effects of social signals in peer recommendation.  ...  Specifically, after controlling for variation due to item content and its position, we find that social signals affect item popularity about half as much as position and content do.  ...  We recorded all actions, including recommendations and URL clicks, and the position of all stories each participant saw. After a vote, the button changed color to indicate the user voted  ... 
arXiv:1410.6744v2 fatcat:2nblseaukfdy3gkhaqxniyhd4m
« Previous Showing results 1 — 15 out of 32,772 results