A Hedging Annotation Scheme Focused on Epistemic Phrases for Informal Language

Abstract

Most existing annotation schemes for hedging were created to aid in the automatic identification of hedges in formal language styles, such as used in scholarly prose. Language with informal tone, typical in much web content, poses a challenge and provides illuminating case studies for the analysis of the use of hedges. We have analysed conversations from a web forum and identified the manners individuals express hedging through expressions which differ slightly regarding to their lexical form from hedges used in formal writing. Based on these observations, we propose an annotation scheme composed of three main categories of hedges where the main class comprises first person epistemic expressions that explicitly note an individual’s involvement in what they express. We provide here an overview of our insights obtained by annotating a dataset of web forum posts according to this scheme. These observations will be useful in the design of automatic methods for the detection of hedges in texts in informal language.