Communications of the ACM
Cumulated gain-based evaluation of IR techniques
ACM Transactions on Information Systems (TOIS)
Optimizing search engines using clickthrough data
Proceedings of the eighth ACM SIGKDD international conference on Knowledge discovery and data mining
Scaling personalized web search
WWW '03 Proceedings of the 12th international conference on World Wide Web
Information Retrieval
Accurately interpreting clickthrough data as implicit feedback
Proceedings of the 28th annual international ACM SIGIR conference on Research and development in information retrieval
Personalizing search via automated analysis of interests and activities
Proceedings of the 28th annual international ACM SIGIR conference on Research and development in information retrieval
Query chains: learning to rank from implicit feedback
Proceedings of the eleventh ACM SIGKDD international conference on Knowledge discovery in data mining
Learning to rank using gradient descent
ICML '05 Proceedings of the 22nd international conference on Machine learning
ACM Transactions on Information Systems (TOIS)
Learning user interaction models for predicting web search result preferences
SIGIR '06 Proceedings of the 29th annual international ACM SIGIR conference on Research and development in information retrieval
Improving web search ranking by incorporating user behavior information
SIGIR '06 Proceedings of the 29th annual international ACM SIGIR conference on Research and development in information retrieval
Evaluating the accuracy of implicit feedback from clicks and query reformulations in Web search
ACM Transactions on Information Systems (TOIS)
Generalized Bradley-Terry Models and Multi-Class Probability Estimates
The Journal of Machine Learning Research
The impact of caching on search engines
SIGIR '07 Proceedings of the 30th annual international ACM SIGIR conference on Research and development in information retrieval
Active exploration for learning rankings from clickthrough data
Proceedings of the 13th ACM SIGKDD international conference on Knowledge discovery and data mining
An experimental comparison of click position-bias models
WSDM '08 Proceedings of the 2008 International Conference on Web Search and Data Mining
Proceedings of the 31st annual international ACM SIGIR conference on Research and development in information retrieval
A user browsing model to predict search engine click data from past observations.
Proceedings of the 31st annual international ACM SIGIR conference on Research and development in information retrieval
Efficient multiple-click models in web search
Proceedings of the Second ACM International Conference on Web Search and Data Mining
Click chain model in web search
Proceedings of the 18th international conference on World wide web
Investigating the effectiveness of clickthrough data for document reordering
ECIR'08 Proceedings of the IR research, 30th European conference on Advances in information retrieval
A statistical model of query log generation
SPIRE'06 Proceedings of the 13th international conference on String Processing and Information Retrieval
Bayesian Browsing Model: Exact Inference of Document Relevance from Petabyte-Scale Data
ACM Transactions on Knowledge Discovery from Data (TKDD)
Learning to re-rank web search results with multiple pairwise features
Proceedings of the fourth ACM international conference on Web search and data mining
Query suggestion by constructing term-transition graphs
Proceedings of the fifth ACM international conference on Web search and data mining
An Online Learning Framework for Refining Recency Search Results with User Click Feedback
ACM Transactions on Information Systems (TOIS)
Why not, WINE?: towards answering why-not questions in social image search
Proceedings of the 21st ACM international conference on Multimedia
Hi-index | 0.00 |
No search engine is perfect. A typical type of imperfection is the preference misalignment between search engines and end users, e.g., from time to time, web users skip higher-ranked documents and click on lower-ranked ones. Although search engines have been aggressively incorporating clickthrough data in their ranking, it is hard to eliminate such misalignments across millions of queries. Therefore, we, in this paper, propose to accompany a search engine with an "always-on" component that reorders documents on a per-query basis, based on user click patterns. Because of positional bias and dependencies between clicks, we show that a simple sort based on click counts (and its variants), albeit intuitive and useful, is not precise enough. In this paper, we put forward a principled approach to reordering documents by leveraging existing click models. Specifically, we compute the preference probability that a lower-ranked document is preferred to a higher-ranked one from the Click Chain Model (CCM), and propose to swap the two documents if the probability is sufficiently high. Because CCM models positional bias and dependencies between clicks, this method readily accounts for many twisted heuristics that have to be manually encoded in sort-based approaches. For this approach to be practical, we further devise two approximation schemes that make online computation of the preference probability feasible. We carried out a set of experiments based on real-world data from a major search engine, and the result clearly demonstrates the effectiveness of the proposed approach.