Abstract

The accuracy of automatic word-class tagging of corpora is tested through a comparison of a manual analysis and the automatic word-class tagging of samples of the occurrences of three high-frequency multifunctional words, as, like, and so, in the Wellington Corpora of Spoken and Written New Zealand English. The results of the comparison show rather high error rates in automatic word-class tagging of these sorts of words.

Download full article (PDF)