Maintaining consistency of monolingual verb entries with interannotator agreement

Resumé

We study the verb throw from the point of the inter-annotator agreement in the Word Sense Disambiguation (WSD) discipline. The most frequently used verbs are often polysemous. This poses a challenge when creating manually annotated data for machine learning, since polysemy threatens inter-annotator agreement. We argue that the classical WSD setup (selecting just one matching sense from a list) is inadequate for semantically complex verbs.