Identification of Topic and Focus in Czech: Evaluation of Manual Parallel Annotations

Abstract

We suppose that the bipartition of the sentence into its Topic and Focus ('aboutness') can be automatically derived from the values of contextual boundness assigned to each node of the dependency tree representing the underlying structure of the sentence. For the testing of this hypothesis, control manual parallel annotations have been carried out.

The principles of the control annotations are described and preliminary results are reported on.

Keywords

• Identification
• Topic
• Focus
• Czech
• Evaluation
• Manual
• Parallel
• Annotations