Combination of Rule-based and Machine Learning for Biomedical Event Extraction

This paper describes the method for biomedical event extraction. The biomedical events occurs in relative to biomedical concepts (objects) as proteins, genes. In this work, we try a hybrid method to identify given event types relative to a given set of proteins in biomedical text. The approach combines rule-based and machine learning. A Set of rules is built based on event triggers, and a set of features is selected to use for machine learning algorithm. Our system consists of four main phases: preprocessing, trigger detection, event detection and post-processing. These phases are developed based on UIMA1 framework. This work is continuous of our work for BioNLP2013 Shared Task2. The final result obtains 36.60 f-score