PIKES is a Java-based suite that extracts knowledge from textual resources. The tool implements a rule-based strategy that reinterprets the output of semantic role labelling (SRL) tools in light of other linguistic analyses, such as dependency parsing or co-reference resolution, thus properly capturing and formalizing in RDF important linguistic aspects such as argument nominalization, frame-frame relations, and group entities.
- Argument nominalization using SRL
- Frame-frame relations extractions
- Entity grouping exploiting linking and co-reference
- Extensible and replaceable NLP pipeline
- Interlinked three-layer representation model exposed as RDF
- Instance RDF triples annotated with detailed information of the mentions (via named graph)
- REST API service included, built on top of Grizzly
- Based on Java 8 and RDFpro
- 2018-05-25 Paper on PSL4EA accepted at ISWC2018! Additional material available here.
- 2018-04-16 Paper on JPARK accepted at IJCAI-ECAI-18! Additional material available here.
- 2017-10-27 Added instructions to use PIKES for batch processing documents.
- 2017-04-18 Added documentation in download page.
- 2016-09-01 Video recording of the ESWC2016 presentation available on videolectures.net
- 2016-08-18 Paper accepted at IEEE TKDE journal!
- 2016-03-18 Added Using PIKES for Information Retrieval section
- 2016-02-23 Paper on using PIKES for Information Retrieval accepted at ESWC 2016 conference!
- 2015-11-24 Paper accepted at SAC 2016 conference, SWA track
- 2015-08-07 Demo accepted at ISWC 2015 conference!