Annotating omission in statement pairs

Abstract

In this piece of industrial application, we focus on the identification of omission in statement pairs for an online news platform. We compare three annotation schemes, namely two crowdsourcing schemes and an expert annotation. The simplest of the two crowdsourcing approaches yields a better annotation quality than the more complex one. We use a dedicated classifier to assess whether the annotators' behaviour can be explained by straightforward linguistic features. However, for our task, we argue that expert and not crowdsourcing-based annotation is the best compromise between cost and quality.