Learning user interaction models for predicting web search result preferences

By Eugene Agichtein, Eric Brill, Susan T. Dumais, Robert J. Ragno. Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, 2006

577 citations recorded by the source corpus.

Subfields: Information Systems

View Paper: https://www.semanticscholar.org/paper/d4753af9568884c1ae9cdcb7617f3c807b22ae61

How researchers see this paper

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

In Information Systems

Citations in the corpus

Time-Aware Click Model (2016, methodology): “DOI: http://dx.doi.org/10.1145/2988230”
Test Collection Based Evaluation of Information Retrieval Systems (2010, unknown): “[1] used query logs to learn how to customize search results for individual users.”
Novelty and diversity in information retrieval evaluation (2008, background): “Relevance judgments inferred from implicit user feedback may not always be accurate [1, 2, 16, 21].”
Analyzing and evaluating query reformulation strategies in web search logs (2009, background): “Several later studies agree that click data are indicators of search result preferences and discuss best methods of analyzing click data [2][12].”
A dynamic bayesian network click model for web search ranking (2009, methodology): “Several machine learning algorithms have been applied to automatically optimize ranking functions [4, 5].”
A user browsing model to predict search engine click data from past observations. (2008, background): “This information can be fed back into the engine, to tune search parameters or even used as direct evidence to influence ranking [2, 10].”
Collaborative Filtering Recommender Systems (2011, background): “[2] is helpful for approximating α in the absence of domain-specific data; their analysis of search query logs shows that 50% of search result clicks are on the top 5 items, making α = 5 a reasonable starting point.”
Characterizing the influence of domain expertise on web search behavior (2009, methodology): “Search result click-through data has been used previously to develop models of user preferences [1].”
Learning to rank for information retrieval (2009, unknown): “• Ground truth mining [3, 68, 105], which targets automatically mining ground truth labels for learning to rank, mainly from click-through logs of search engines.”
Understanding web browsing behaviors through Weibull analysis of dwell time (2010, methodology): “[2, 3] demonstrate that user browsing features, a major component of which is Web page dwell time, significantly improve the retrieval performance of a competitive search engine, even with the presence of other important features such as BM25 and search-result clickthrough.”
Finding high-quality content in social media (2008, methodology): “Interestingly, including all derived features similar to those described in [1] actually degrades performance, indicating overfitting when relying on usage statistics alone without the benefit of other forms of user feedback.”
Answer Interaction in Non-factoid Question Answering Systems (2019, background): “User interaction with SERPs has been widely studied using search logs that contain clicks and query reformulations [2, 23].”
Evaluating the accuracy of implicit feedback from clicks and query reformulations in Web search (2007, methodology): “, [Fuhr 1989; Bartell et al. 1994]). While Cohen et al. [1999] discussed the use of clickthrough data, they derived the data for their experiments from explicit judgments.”
How does clickthrough data reflect retrieval quality? (2008, background): “Others who have studied usage-based retrieval evaluation include [4, 13, 1, 2, 9, 11, 18].”
Designing Ranking Systems for Hotels on Travel Search Engines by Mining User-Generated and Crowdsourced Content (2012, background): “Ofcourse,diversity of results is a well-known factor of user satisfaction in Web search (Agichtein et al. 2006).”
An experimental comparison of click position-bias models (2008, background): “A cascade model, where users view results from top to bottom and leave as soon as they see a worthwhile document, is our best explanation for position bias in early ranks.”