Figure 1.

Event extraction. In most tasks named entities are given (A). Sentences are parsed (B) to produce a dependency parse. Entities not given are predicted through trigger detection (C). Edge detection predicts event arguments between entities (D) and unmerging creates events (E). Finally, event modality is predicted (F). When the graph is converted to the Shared Task format, site arguments are paired with core arguments that have the same target protein.

Bj√∂rne et al. BMC Bioinformatics 2012 13(Suppl 11):S4   doi:10.1186/1471-2105-13-S11-S4