I. Gurevych

Publications (3)0 Total impact

• Source

  Available from: Torsten Zesch

  Conference Proceeding: Link Discovery: A Comprehensive Analysis

  N. Erbs, T. Zesch, I. Gurevych
  [show abstract] [hide abstract]
  ABSTRACT: We present a comprehensive analysis of link discovery approaches. We classify them with regard to the type of knowledge being used, and identify three commonly used sources of knowledge: The text of a document, the document title, and already existing links. We analyze the influence of the knowledge source as well as of the amount of training data used. Results show that the link-based approach performs best if the amount of training data is huge. In a more realistic setting with fewer training data, the text-based approach yields better results.
  Semantic Computing (ICSC), 2011 Fifth IEEE International Conference on; 10/2011
• Source

  Available from: Torsten Zesch

  Conference Proceeding: Using Similarity Measures for Context-Aware User Interfaces

  M. Hartmann, T. Zesch, M. Muhlhauser, I. Gurevych
  [show abstract] [hide abstract]
  ABSTRACT: Context-aware user interfaces facilitate the user interaction by suggesting or prefilling data derived from the userpsilas current context. This raises the problem of mapping context information to input elements in the user interface. We address this problem for web applications by (i) automatically extracting a textual representation of their input elements, and by (ii) mapping context information to them using these textual representations. In this paper, we present an approach for the representation extraction task that outperforms existing ones, and we explore the potential of similarity measures for the context mapping task.
  Semantic Computing, 2008 IEEE International Conference on; 09/2008
• Source

  Available from: Niklas Jakob

  Conference Proceeding: A Comparative Study of Feature Extraction Algorithms in Customer Reviews

  L. Ferreira, N. Jakob, I. Gurevych
  [show abstract] [hide abstract]
  ABSTRACT: The paper systematically compares two feature extraction algorithms to mine product features commented on in customer reviews. The first approach [17] identifies candidate features by applying a set of POS patterns and pruning the candidate set based on the log likelihood ratio test. The second approach [11] applies association rule mining for identifying frequent features and a heuristic based on the presence of sentiment terms for identifying infrequent features. We evaluate the performance of the algorithms on five product specific document collections regarding consumer electronic devices. We perform an analysis of errors and discuss advantages and limitations of the algorithms.
  Semantic Computing, 2008 IEEE International Conference on; 09/2008