Abstract
In this work, we have developed an open source, portable toolkit for detecting negation cues and their scope in natural language. Our tool is designed with a two-phase architecture, where cue detection and scope resolution are solved using two independent machine learning classifiers. In our implementation, we have built upon the best practices from previous work, in terms of feature design, machine learning algorithms, datasets and evaluation methods, and built the entire system from scratch through large- scale experiments to assess the utility of features and different classifiers.