next up previous contents
Next: Methodology: Statistics underlying the Up: Tagsets and Taggers Interaction Previous: Other taggers

Proposals for practical experiments


In general, taggers are evaluated by computing the amount of tagging errors, for example by comparing an automatically tagged corpus to a manually tagged corpus. This error rate, however, does not only depend on the tagging method or tagset, but also on the ambiguity rate and accuracy of the lexicon that is used. Test results also may vary for different types and sizes of the test corpus.