Getting at the Cognitive Complexity of Linguistic Metadata Annotation – A Pilot Study Using Eye-Tracking

We report on an experiment where the decision behavior of annotators issuing linguistic metadata is observed with an eyetracking device. As experimental conditions we consider the role of textual context and linguistic complexity classes. Still preliminary in nature, our data suggests that semantic complexity is much harder to deal with than syntactic one, and that full-scale textual context is negligible for annotation, with the exception of semantic high-complexity cases. We claim that such observational data might lay the foundation for empirically grounded annotation cost models and the design of cognitively adequate annotation user interfaces.