Integrating Shallow and Deep NLP for Information Extraction

Feiyu Xu; Hans-Ulrich Krieger

In: Proceedings of RANLP 2003. International Conference on Recent Advances in Natural Language Processing (RANLP), 9/2003.


This paper describes a novel approach to information extraction by developing strategies for combining techniques from shallow and deep NLP. We propose a hybrid template filling strategy, which employs shallow partial syntactic analysis for extracting local domain-specific relations and uses predicate-argument structures delivered by deep full-sentence analysis for extracting relations triggered by verbs. Heuristics have been developed for calling deep NLP on demand. The initial evaluation shows that the integration of deep analysis improves the performance of the scenario template-generation task.

WHIES.pdf (pdf, 186 KB )

Deutsches Forschungszentrum für Künstliche Intelligenz
German Research Center for Artificial Intelligence