Rule-based chunking of semantic roles for Turkish

In our work, we approached to semantic role labeling from a different angle. Contrary to related works, which focused on determining single role like noun phrase or predicate, we worked on all of the roles. We claim that, morphological analysis of a word and its context can be useful for semantic role labeling task. For that, we first determine the possible semantic chunk boundaries by examining the morphological analysis of words and their contexts. For further improvement in determining the boundaries, we do the first process with the combination of the morphological analysis and the boundary output from the first pass. We use these boundaries to create semantic chunks and labeled them according to their content.