Guided Linguistic Annotation of Argumentation through Visual Analytics


We present a mixed-initiative approach to interactive annotation of argumentation, a typically time-consuming manual task. Our system facilitates the process by suggesting which fragments of text to annotate next. Suggestions are sourced from pre-annotations and user-preferences that are learned over time. Unused suggestions decay over time, reducing the amount of necessary interactions, while providing additional training data to the system. We show the effectiveness of the system for argument annotation according to Inference Anchoring Theory. The duality of suggestion sources and novel approach to suggestion decay are broadly applicable in linguistic annotation

COMMA Workshop on Argument Visualization (ArgVis)