Centrum Wiskunde & Informatica: Modelling Complex Relevance Spaces with Copulas

<!DOCTYPE html> <html lang="en"> <head> <title> Centrum Wiskunde & Informatica: Modelling Complex Relevance Spaces with Copulas </title> <meta http-equiv="X-UA-Compatible" content="IE=edge"/> <meta http-equiv="Content-Type" content="text/html; charset=UTF-8"/> <meta name="viewport" content="width=device-width, initial-scale=1.0"/> <meta name="citation_title" content="Modelling Complex Relevance Spaces with Copulas"/> <meta name="citation_author" content="Eickhoff, Carsten"/> <meta name="citation_author" content="de Vries, Arjen"/> <meta name="citation_date" content="2014/11/01"/> <meta name="citation_publication_date" content="2014/11/01"/> <meta name="citation_online_date" content="2014/11/01"/> <meta name="citation_abstract" content="Modern relevance models consider a wide range of criteria in order to identify those documents that are expected to satisfy the user's information need. With growing dimensionality of the underlying relevance spaces the need for sophisticated score combination and estimation schemes arises. In this paper, we investigate the use of copulas, a model family from the domain of robust statistics, for the formal estimation of the probability of relevance in high-dimensional spaces. Our experiments are based on the MSLR-WEB10K and WEB30K datasets, two annotated, publicly available samples of hundreds of thousands of real Web search impressions, and suggest that copulas can significantly outperform linear combination models for high-dimensional problems. Our models achieved a performance on par with that of state-of-the-art machine learning approaches." /> <meta name="citation_doi" content="10.1145/2661829.2661925" /> <meta name="citation_pdf_url" content="https://ir.cwi.nl/pub/23442/23442D.pdf"/> <link rel="alternate" type="application/rdf+xml" title="VIVO" href="https://ir.cwi.nl/pub/23442/rdf"/> <link rel="repository" href="https://ir.cwi.nl/"/> <link rel="me" href="https://ir.cwi.nl/pub/23442/"/> <link rel="view" href="https://ir.cwi.nl/pub/23442"/> <link rel="shortcut icon" type="image/x-icon" href="/static/img/favicon.ico"/> <link rel="stylesheet" type="text/css" href="/static/css/bundle-1.css"/> <script type="text/javascript" src="/static/js/bundle-1.js"></script>   </head> <body> <div id="wrap"> <header> <div class="navbar" style="margin-bottom:0"> <div class="navbar-inner"> <div class="container"> <a class="brand" style="padding:10px;width:200px" href="https://ir.cwi.nl/"> <img alt="university website" style="width:100%" src="https://ir.cwi.nl/cwi.png"/> </a> <a class="hidden" href="https://ir.cwi.nl/year">Publications by Year</a> <ul class="nav"> </ul> <ul class="nav pull-right"> <li id="fat-menu" class="dropdown"> <a class="dropdown-toggle" href="https://ir.cwi.nl/login?next=/pub/23442"> <span class="icon-signin"></span> sign in </a> </li> </ul> </div> </div> </div> </header> <div id="breadcrumb" class="row-fluid"> <div class="span12"> <ul class="breadcrumb"> <li> <a href="https://ir.cwi.nl/org/24">Human-Centered Data Analytics</a> <span class="divider">/</span> </li> <li class="active">In Proceedings</li> <li class="pull-right"> <div class="input-append search-box" style="margin-top:-5px"> <label for="search-query-box" style="display:none">Search:</label> <input type="text" class="input-large search-input" id="search-query-box" style="width:250px" value="" name="query"/> <a class="btn btn-primary search-button" style="border-bottom:1px solid rgba(0, 0, 0, 0.25)" href="https://ir.cwi.nl/search#"> <span class="icon-search"></span> Search </a> </div> </li> </ul> </div> </div> <div class="row-fluid"> <div id="main-left" class="span1"> </div> <div id="main" class="span10"> <div id="content"> <div class="page-header"> <div class="clearfix"> <p class="lead author-list pull-left"> <span style=""> <a href="https://ir.cwi.nl/ppl/6102"> <span>C. Eickhoff (Carsten)</span></a><span> and </span> </span> <span style=""> <a href="https://ir.cwi.nl/ppl/1"> <span>A.P. de Vries (Arjen)</span></a> </span> <br class="clearfix"/> </p> <p class="lead pub_type pull-right"> <span>2014-11-01</span> </p> </div> <h1>Modelling Complex Relevance Spaces with Copulas</h1> <h2 style="display:none">Publication</h2> <h3 style="display:none">Publication</h3> <div> <em> Presented at the <span>ACM Conference on Information and Knowledge Management </span> </em> </div> </div> <div class="row-fluid"> <div class="span8">  <div class="abstract">Modern relevance models consider a wide range of criteria in order to identify those documents that are expected to satisfy the user's information need. With growing dimensionality of the underlying relevance spaces the need for sophisticated score combination and estimation schemes arises. In this paper, we investigate the use of copulas, a model family from the domain of robust statistics, for the formal estimation of the probability of relevance in high-dimensional spaces. Our experiments are based on the MSLR-WEB10K and WEB30K datasets, two annotated, publicly available samples of hundreds of thousands of real Web search impressions, and suggest that copulas can significantly outperform linear combination models for high-dimensional problems. Our models achieved a performance on par with that of state-of-the-art machine learning approaches.</div>  <div id="publication-metadata"> <table class="table"> <thead> <tr><th colspan="2" class="publiation-metadata-header">Additional Metadata</th></tr> </thead> <tbody> <tr> <td style="min-width:120px"> <span class="publication-metadata-key">THEME</span> </td> <td> <span class="publication-metadata-value" key="THEME">Information (theme 2)</span> </td> </tr> <tr> <td style="min-width:120px"> <span class="publication-metadata-key">Publisher</span> </td> <td> <span>ACM</span> </td> </tr> <tr> <td style="min-width:120px"> <span class="publication-metadata-key">Persistent URL</span> </td> <td> <a href="https://doi.org/10.1145/2661829.2661925">doi.org/10.1145/2661829.2661925</a> </td> </tr> <tr> <td style="min-width:120px"> <span class="publication-metadata-key">Conference</span> </td> <td> <span>ACM Conference on Information and Knowledge Management </span> </td> </tr> <tr> <td class="publication-metadata-key"> Organisation </td> <td> <a href="https://ir.cwi.nl/org/24">Human-Centered Data Analytics</a> </td> </tr> <tr> <td><span class="publication-metadata-key">Citation</span><br/> <div class="btn-group dropup" id="citation-menu"> <a class="btn dropdown-toggle btn-small" id="citation-menu-button" data-toggle="dropdown" href="#"> APA Style <span class="icon-chevron-up"></span> </a> <ul class="dropdown-menu"> <li> <a href="#" class="citation-style-pub" data-citation-style="aaa"> AAA Style </a> </li> <li> <a href="#" class="citation-style-pub" data-citation-style="apa"> APA Style </a> </li> <li> <a href="#" class="citation-style-pub" data-citation-style="cell"> Cell Style </a> </li> <li> <a href="#" class="citation-style-pub" data-citation-style="chicago"> Chicago Style </a> </li> <li> <a href="#" class="citation-style-pub" data-citation-style="harvard"> Harvard Style </a> </li> <li> <a href="#" class="citation-style-pub" data-citation-style="ieee"> IEEE Style </a> </li> <li> <a href="#" class="citation-style-pub" data-citation-style="mla"> MLA Style </a> </li> <li> <a href="#" class="citation-style-pub" data-citation-style="nature"> Nature Style </a> </li> <li> <a href="#" class="citation-style-pub" data-citation-style="vancouver"> Vancouver Style </a> </li> <li> <a href="#" class="citation-style-pub" data-citation-style="american-institute-of-physics"> American-Institute-of-Physics Style </a> </li> <li> <a href="#" class="citation-style-pub" data-citation-style="council-of-science-editors"> Council-of-Science-Editors Style </a> </li> <li class="divider"/> <li> <a href="#" class="citation-format-pub" data-citation-format="bibtex"> BibTex Format </a> </li> <li> <a href="#" class="citation-format-pub" data-citation-format="endnote"> Endnote Format </a> </li> <li> <a href="#" class="citation-format-pub" data-citation-format="ris"> RIS Format </a> </li> <li> <a href="#" class="citation-format-pub" data-citation-format="csl"> CSL Format </a> </li> <li> <a href="#" class="citation-format-pub" data-citation-format="dois only"> DOIs only Format </a> </li> </ul> </div> </td> <td id="citation-text"> <div class="csl-bib-body"><div class="csl-entry">Eickhoff, C., & de Vries, A. (2014). Modelling Complex Relevance Spaces with Copulas. In <em>Proceedings of the 23rd ACM International Conference on Conference on Information and Knowledge Management</em> (pp. 1831–1834). ACM. doi:10.1145/2661829.2661925</div></div> </td> </tr> </tbody> </table> </div> </div> <div class="span4"> <ul class="publication-identifiers unstyled"> <li> <a href="https://doi.org/10.1145/2661829.2661925"><span class="icon-link"></span> View at Publisher</a> </li> </ul> <hr style="margin:5px 0px 5px 0px;"/> <div class="publication-downloads"> <legend> <a href="https://ir.cwi.nl/pub/23442/23442D.pdf"> <span class="icon-download-alt"></span> <span class="download-link-text"> Free Full Text </span> <small class="download-link-info"> ( Final Version , 886kb ) </small> </a> </legend> <a href="https://ir.cwi.nl/pub/23442/23442D.pdf"> <img src="https://ir.cwi.nl/pub/23442/cover/thumb_256.jpg" alt="cover" id="primary_asset_cover_img" class="shadow-book"/> </a> </div> <div> <br/> <table class="table" id="additional_assets_table"> <thead> <tr><th colspan="2">Additional Files</th></tr> </thead> <tbody> <tr> <td> <span class="icon-book"></span> <a href="https://ir.cwi.nl/pub/23442/Fulltext" target="_blank"> <span>Fulltext</span> </a> <span class="pull-right"> Final Version </span> </td> </tr> <tr> <td> <span class="icon-external-link"></span> <a href="https://ir.cwi.nl/pub/23442/Publisher%20Version" target="_blank"> <span>Publisher Version</span> </a> <span class="pull-right"> </span> </td> </tr> </tbody> </table> </div> <br/> </div> </div> </div> </div> <div id="main-right" class="span1"> </div> </div> </div> <footer class="footer row-fluid"> <div class="span10 offset1"> <div class="row-fluid"> <div class="span3"> <a> <img alt="university website" src="https://ir.cwi.nl//cwi.png"/> </a> <ul class="unstyled"> </ul> </div> <div class="span3"> <ul class="unstyled"> <li> <div> <h4>Address</h4> <ul class="unstyled" style="margin-left:0"> <li> <div class="lead" style="margin:4px"> <a href="http://www.cwi.nl"> Centrum Wiskunde & Informatica </a> </div> </li> <li> <div class="lead" style="margin:4px"> <a href="http://www.cwi.nl"> Science Park 123 | 1098 XG Amsterdam | the Netherlands </a> </div> </li> </ul> </div> </li> </ul> </div> <div class="span3"> <ul class="unstyled"> <li> <div> <h4>CWI researchers</h4> <ul class="unstyled" style="margin-left:0"> <li> <div class="lead" style="margin:4px"> <a href="/page/1"> <span class="icon-file"></span> Register Your Work </a> </div> </li> </ul> </div> </li> <li> <div> <h4>Questions or comments?</h4> <ul class="unstyled" style="margin-left:0"> <li> <div class="lead" style="margin:4px"> <a href="mailto:repository@cwi.nl"> <span class="icon-envelope"></span> repository@cwi.nl </a> </div> </li> </ul> </div> </li> </ul> </div> </div> </div> <div class="span1"> <a href="http://www.artudis.com" class="pull-right" style="margin: 170px 20px 10px 10px"> <img src="/static/img/logo/artudis_small.png" width="100" alt="artudis website"/> </a> </div> </footer> <div id="modal-workflow-editor" class="modal hide fade"> <div class="modal-header"> <button type="button" class="close" data-dismiss="modal" aria-hidden="true">×</button> <h3 id="modal-workflow-header">Workflow</h3> </div> <div id="modal-workflow-form"></div> </div> <div id="modal-iframe" style="width:800px" class="modal hide fade"> <div class="modal-header"> <button type="button" class="close" data-dismiss="modal" aria-hidden="true">×</button> <h3 id="modal-iframe-header">Workflow</h3> </div> <div> <iframe id="modal-iframe-body" src="" style="width:100%;height:500px" frameborder="0" ></iframe> </div> </div> <div id="add-content-modal" class="modal hide fade"> <div class="modal-header"> <button type="button" class="close" data-dismiss="modal" aria-hidden="true">×</button> <h3 id="modal-edit-header">Add Content</h3> </div> <div class="modal-body"> <div class="btn-group btn-group-vertical" style="width:100%;max-height:600px;overflow:auto"> <br/> <a class="btn btn-large" style="margin:auto;width:300px" href="https://ir.cwi.nl/user/add">User</a> <a class="btn btn-large" style="margin:auto;width:300px" href="https://ir.cwi.nl/pub/add">Publication</a> <a class="btn btn-large" style="margin:auto;width:300px" href="https://ir.cwi.nl/ppl/add">Person</a> <a class="btn btn-large" style="margin:auto;width:300px" href="https://ir.cwi.nl/org/add">Organisation</a> <a class="btn btn-large" style="margin:auto;width:300px" href="https://ir.cwi.nl/col/add">Collection</a> </div> </div> <div class="modal-footer"> <a href="#" class="btn btn-primary" data-dismiss="modal">Close</a> </div> </div> </body> </html>

CINXE.COM

Centrum Wiskunde & Informatica: Modelling Complex Relevance Spaces with Copulas