Multilabel Learning for Text-WebResource Classification

Over multiple years, a large amount of web-portals developed, providing their content to different groups of users. As the several content-providers usually focus on specialized fields of knowledge, they developed schemes of resource-classification in terms of content-management-systems, underlying the portals to manage the delivery of new published items, which are often inspired by the traditional news-publishing-schemes known from the field of journalism. As users usually read across the internet over multiple sources, they develop over browsing- and reading-sessions an own view, which content-providers share the most relevant content, such that users revisit the portals again when looking for new up-to-date content. To manage their known and read resources and for building an own library-catalogue-scheme, users apply the technique of storing a web-resource shortcut-descriptor-item as bookmark to their 'bookmark-knowledge-library', organized usually flat but partially hierarchical in bookmark folders. Over sessions, it can be observed, that users build their own organization and classification scheme for their shortcut-library, mostly build related to the processed content.

Autorentext

Bastian Christoph, BSc Informatik:Student at Computer Science Department, TU Darmstadt;Student Research Assistant:TextCollectionProcessing (DataScience){2011; UKP Group; UserManual for TextSummarization; TextNews Dataset 'DUC 2001-2005'},{2009-2011; KOM Group; Extraction of 'Hyponym-Hypernym' semantic-relation wordpair-instances from Wikipedia}