Assessing the validity and reliability of Turkish Discourse Bank (TDB 1.0)

Manually annotated linguistic corpora record language users’ intuitions electronically. Reusability of annotated corpora depends on adopting an annotation approach which aspires to be linguistically sound and reliable. The major goal of this study is to present an overall assessment of the validity and reliability of Turkish Discourse Bank (TDB 1.0), the first discourse-annotated corpus of Turkish, via two approaches referred to as the Overall Approach and the Common Approach. The chapter discusses the results both quantitatively and qualitatively and hopes to contribute to an evolving effort that brings together linguists, corpus linguists and computer scientists to move towards good practices in annotation science from the viewpoint of Turkish discourse.