Automated construction of a large semantic network of related terms for domain-specific modeling

  • Authors:
  • Henning Agt;Ralf-Detlef Kutsche

  • Affiliations:
  • Database Systems and Information Management Group DIMA, Technische Universität Berlin, Berlin, Germany;Database Systems and Information Management Group DIMA, Technische Universität Berlin, Berlin, Germany

  • Venue:
  • CAiSE'13 Proceedings of the 25th international conference on Advanced Information Systems Engineering
  • Year:
  • 2013

Quantified Score

Hi-index 0.00

Visualization

Abstract

In order to support the domain modeling process in model-based software development, we automatically create large networks of semantically related terms from natural language. Using part-of-speech tagging, lexical patterns and co-occurrence analysis, and several semantic improvement algorithms, we construct SemNet, a network of approximately 2.7 million single and multi-word terms and 37 million relations denoting the degree of semantic relatedness. This paper gives a comprehensive description of the construction of SemNet, provides examples of the analysis process and compares it to other knowledge bases. We demonstrate the application of the network within the Eclipse/Ecore modeling tools by adding semantically enhanced class name autocompletion and other semantic support facilities like concept similarity.