Relational Learning of Pattern-Match Rules for Information Extraction

Information extraction is a form of shallow text processingthat locates a specified set of relevant items in a naturaManguage document. Systems for this task require significant domain-specific knowledgeand are time-consumingand difficult to build by hand, makingthem a good application for machinelearning. Wepresent a system, RAPIER, that uses pairs of sample documentsand filled templates to inducepattern-matchrules that directly extract fillers for the slots in the template. RAPIER employs a bottom-up learning algorithm whichincorporates techniques fromseveral inductive logic programming systems and acquires unboundedpatterns that include constraints on the words, part-of-speech tags, and semantic classes present in the filler andthe surroundingtext. We present encouraging experimental results on two domains.

Authors

• Mary Elaine Califf
• Illinois State University
• Raymond J. Mooney
• University of Texas at Austin

Keywords

No keywords are indexed for this paper.