Babarczy, A., J. Carroll and G. Sampson (2001) `Annotator error rates for part-of-speech tagging'. Presented at Workshop Linguistically Interpreted Corpora (LINC-2001), Leuven, Belgium.
Part-of-speech tagging is an important and ubiquitous type of corpus annotation. We report the results of experiments establishing an upper bound for tagging accuracy, and typical annotator error rates under various post-editing scenarios. The work has significant consequences for quality assurance of annotated corpora, reporting of error rates for automatic tagging, and annotation workflow practices.