Generic Relation Extraction

1 minute read

webLyzard is a platform for the automated analysis of content, navigational system and interface design of networked information systems. Through structural and textual metrics, webLyzard determines success factors and uncovers weaknesses of deployed system by comparing large samples of Web sites across region and industry. Currently, the project mirrors and analyses more 500,000 sites per week.

Many current research projects, such as IDIOM, RAVEN, and the Election 2008 Monitor depend on applying state of the art pre-processing, statistical, and natural language detection methods to be applied to this corpus.

Processing such amounts of documents requires appropriate techniques which scale well even under Web-scale conditions (Cafarella et al.). Open relation extraction approaches such as the one introduced in Banko and Etzioni address this issue.

The goals of this thesis are to

  1. investigate the state of the art in Web-scale relation extraction techniques,
  2. implement a research prototype in Java or Python:
    • capable of extracting relations from documents in linear time.
    • domain independent


  1. introduction and problem definition
  2. theoretical considerations
  3. state of the art
  4. implementation
    1. requirement analysis
    2. system description
    3. java or python prototype
  5. evaluation
  6. outlook and conclusions


Essential Literature

  • [Banko:2008] Banko, Michele and Etzioni, Oren (2008). The Tradeoffs Between Open and Traditional Relation Extraction, Proceedings of ACL-08: HLT, Association for Computational Linguistics, pages 28–36
  • [cafarella2008] Cafarella, Michael J., Madhavan, Jayant and Halevy, Alon (2008). Web-scale extraction of structured data, SIGMOD Rec., ACM, pages 55–61, 37(4)
  • [zelenko2003] Zelenko, Dmitry, Aone, Chinatsu and Richardella, Anthony (2003). Kernel methods for relation extraction, J. Mach. Learn. Res., MIT Press, pages 1083–1106
  • [Nicolae2007] Nicolae, Cristina, Nicolae, Gabriel and Harabagiu, Sanda (2007). UTD-HLT-CG: Semantic Architecture for Metonymy Resolution and Classification of Nominal Relations, Proceedings of the Fourth International Workshop on Semantic Evaluations (SemEval-2007), Association for Computational Linguistics, pages 454–459

Related Literature

  • [cimiano2007] Cimiano, Philipp and Wenderoth, Johanna (2007). Automatic Acquisition of Ranked Qualia Structures from the Web, ACL
  • [zhu2009] Zhu, Jun, Nie, Zaiqing, Liu, Xiaojiang, Zhang, Bo and Wen, Ji-Rong (2009). StatSnowball: a statistical approach to extracting entity relationships, WWW ‘09: Proceedings of the 18th international conference on World wide web, ISBN: 978-1-60558-487-4, ACM, pages 101–110