Active exploration for learning rankings from clickthrough data
F. Radlinski, and T. Joachims. Proceedings of the 13th ACM SIGKDD international conference on Knowledge discovery and data mining, page 570--579. San Jose, California, USA, ACM, (2007)
DOI: 10.1145/1281192.1281254
Abstract
We address the task of learning rankings of documents from search enginelogs of user behavior. Previous work on this problem has relied onpassively collected clickthrough data. In contrast, we show that anactive exploration strategy can provide data that leads to much fasterlearning. Specifically, we develop a Bayesian approach for selectingrankings to present users so that interactions result in more informativetraining data. Our results using the TREC-10 Web corpus, as well assynthetic data, demonstrate that a directed exploration strategy quicklyleads to users being presented improved rankings in an online learningsetting. We find that active exploration substantially outperformspassive observation and random exploration.
%0 Conference Paper
%1 radlinski_active_2007
%A Radlinski, Filip
%A Joachims, Thorsten
%B Proceedings of the 13th ACM SIGKDD international conference on Knowledge discovery and data mining
%C San Jose, California, USA
%D 2007
%I ACM
%K activeexploration,clickthroughdata,learningtorank,websearch
%P 570--579
%R 10.1145/1281192.1281254
%T Active exploration for learning rankings from clickthrough data
%U http://portal.acm.org/citation.cfm?id=1281192.1281254&coll=GUIDE&dl=GUIDE&CFID=7216732&CFTOKEN=41760401
%X We address the task of learning rankings of documents from search enginelogs of user behavior. Previous work on this problem has relied onpassively collected clickthrough data. In contrast, we show that anactive exploration strategy can provide data that leads to much fasterlearning. Specifically, we develop a Bayesian approach for selectingrankings to present users so that interactions result in more informativetraining data. Our results using the TREC-10 Web corpus, as well assynthetic data, demonstrate that a directed exploration strategy quicklyleads to users being presented improved rankings in an online learningsetting. We find that active exploration substantially outperformspassive observation and random exploration.
%@ 978-1-59593-609-7
@inproceedings{radlinski_active_2007,
abstract = {We address the task of learning rankings of documents from search enginelogs of user behavior. Previous work on this problem has relied onpassively collected clickthrough data. In contrast, we show that anactive exploration strategy can provide data that leads to much fasterlearning. Specifically, we develop a Bayesian approach for selectingrankings to present users so that interactions result in more informativetraining data. Our results using the TREC-10 Web corpus, as well assynthetic data, demonstrate that a directed exploration strategy quicklyleads to users being presented improved rankings in an online learningsetting. We find that active exploration substantially outperformspassive observation and random exploration.},
added-at = {2009-03-05T08:49:43.000+0100},
address = {San Jose, California, USA},
author = {Radlinski, Filip and Joachims, Thorsten},
biburl = {https://www.bibsonomy.org/bibtex/28b46552dea22b86815e4804b084dbe9a/bcao},
booktitle = {Proceedings of the 13th ACM SIGKDD international conference on Knowledge discovery and data mining},
doi = {10.1145/1281192.1281254},
interhash = {d69f600db696241b14a2b66daf920f04},
intrahash = {8b46552dea22b86815e4804b084dbe9a},
isbn = {978-1-59593-609-7},
keywords = {activeexploration,clickthroughdata,learningtorank,websearch},
pages = {570--579},
publisher = {ACM},
timestamp = {2009-03-05T08:49:43.000+0100},
title = {Active exploration for learning rankings from clickthrough data},
url = {http://portal.acm.org/citation.cfm?id=1281192.1281254\&coll=GUIDE\&dl=GUIDE\&CFID=7216732\&CFTOKEN=41760401},
year = 2007
}