Next: Resources and tools
Up: Test setups and objectives
Previous: Tagset Evaluation
Differences between training texts and test texts:
- it is assumed that certain text types (e.g. news vs. laws vs.
technical prose) differ with respect to the distribution of syntactic
constructions;
- these text-typical differences may lead to (slightly?) different
statistical models, for different texts.
The objective is to find ways of assessing the impact of such
differences, when they occur between training and test texts.