Influence of preprocessing on dependency syntax annotation: speed and agreement
Arne Skjærholt · Linguistic Annotation Workshop · 2013
When creating a new resource, preprocessing the source texts before annotation is both ubiquitous and obvious. How the preprocessing affects the annotation effort for various tasks is for the most part an open question, however. In this paper, we study the effects of preprocessing on the annotation of dependency corpora and how annotation speed varies as a function of the quality of three different parsers and compare with the speed obtained when starting from a least-processed baseline. We also present preliminary results concerning the effects on agreement based on a small subset of sentences that have been doubly-annotated. 1