This release contains the source code used for Chapter 5 of the PhD thesis "Global Inference and Local Syntax Representations for Event Extraction". It contains the implementation of the modular, graph-based event extractor described there.

Furthermore, the release contains five models produced with the Graph Convolution Network and an repeated negative undersampling probability of 0.9. These five models can either be used alone, or, as intended, as part of a boostrap aggregating implementation of the event extractor.

For more information, see the README.md.