Charles Explorer logo
🇬🇧

Analyzing the Most Common Errors in the Discourse Annotation of the Prague Dependency Treebank

Publication at Faculty of Mathematics and Physics |
2012

Abstract

We present an analysis of the inter-annotator discrepancies of the Czech discourse annotation in the Prague Dependency Treebank 2.0. Having finished the annotation of the inter-sentential semantic discourse relations with explicit connectives in the treebank, we report now on the results of the evaluation of the parallel (double) annotations, which is an important step in the process of checking the quality of the data.

After we shortly describe the annotation and the method of the inter-annotator agreement measurement, we present the results of the measurement and, most importantly, we classify and analyze the most common types of annotators’ disagreement.