2014

referat

angielski

EN In this paper we present a rule-based method of event extraction from the natural language. We use the Stanford dependency parser in order to build a relation graph of elements from input text. This structure along with serialized extraction frames is converted into a set of facts. We describe a process of crea-tion of application of rules, which aims to match elements from the text with corresponding slots in the extraction frames. A possible match is derived by the comparison of verbal phrases from the text with lexicalizations of anchors (con-stituting the most important part of each frame) stored in an ontology. The rest of the extraction frame is filled with other elements of the dependency graph, with regard to their semantic type (determined by lexicalizations of allowed types defined in frames and ontology) and their grammatical properties. We describe conversions required to create a consistent knowledge base of text phrases, clas-sification of semantic types and instantiated slots from the extraction frames. We use the Drools engine in order to extract events from such a knowledge base.

1 - 13

Proceedings of the RuleML 2014 Challenge and the RuleML 2014 Doctoral Consortium hosted by the 8th International Web Rule Symposium (RuleML 2014)

RuleML 2014 Challenge and the RuleML 2014 Doctoral Consortium hosted by the 8th International Web Rule Symposium (RuleML 2014), 18-20.08.2014, Prague, Czech Republic