Browsing Centre for Research in Distributed Technologies (CREDIT) by Authors
Combining cognitive and system-oriented approaches for designing IR user interfacesFuhr, Norbert; Jordan, Matthias; Frommholz, Ingo; University of Duisburg-Essen, Germany (Gesellschaft für Informatik e.V., 2008)
Determining the polarity of postings for discussion searchFrommholz, Ingo; Lechtenfeld, Marc; University of Duisburg-Essen, Germany (Gesellschaft für Informatik e.V., 2008)When performing discussion search it might be desirable to consider non-topical measures like the number of positive and negative replies to a posting, for instance as one possible indicator for the trustworthiness of a comment. Systems like POLAR are able to integrate such values into the retrieval function. To automatically detect the polarity of postings, they need to be classified into positive and negative ones w.r.t.\ the comment or document they are annotating. We present a machine learning approach for polarity detection which is based on Support Vector Machines. We discuss and identify appropriate term and context features. Experiments with ZDNet News show that an accuracy of around 79\%-80\% can be achieved for automatically classifying comments according to their polarity.
Multi-facet classification of e-mails in a helpdesk scenarioBeckers, Thomas; Frommholz, Ingo; Bonning, Ralf; University of Duisburg-Essen, Germany; University of Glasgow; d.velop AG (Gesellschaft für Informatik e.V., 2009)Helpdesks have to manage a huge amount of support requests which are usually submitted via e-mail. In order to be assigned to experts e ciently, incoming e-mails have to be classi- ed w. r. t. several facets, in particular topic, support type and priority. It is desirable to perform these classi cations automatically. We report on experiments using Support Vector Machines and k-Nearest-Neighbours, respectively, for the given multi-facet classi - cation task. The challenge is to de ne suitable features for each facet. Our results suggest that improvements can be gained for all facets, and they also reveal which features are promising for a particular facet.