I've almost completed writing a new part-of-speech tagging scheme. My problem is how to evaluate how effective it is at tagging. What's the best way of doing this at the moment? I have a copy of the Susanne Corpus, and I'll definitely use that in my experiments. What other tagged corpora are out there? (Either public domain or can be purchased). Are there any other good methods of evaluating how good a tagger is? Bill Teahan