A Graphical Interface for Automatic Error Mining in Corpora

We present an error mining tool that is de-signed to help human annotators to find errors and inconsistencies in their anno-tation. The output of the underlying al-gorithm is accessible via a graphical user interface, which provides two aggregate views: a list of potential errors in con-text and a distribution over labels. The user can always directly access the ac-tual sentence containing the potential er-ror, thus enabling annotators to quickly judge whether the found candidate is in-deed incorrectly labeled.