Coreference Resolution for Russian: Taking Stock and Moving Forward

Coreference resolution is targeted at collecting all mentions of an entity within a text into a single cluster. Various approaches for this task have already been proposed but the problem is still challenging. RU-EVAL-2014, an evaluation of anaphora and coreference resolution, renewed interest for this task for the Russian language. We try to follow this trend; our work can be viewed as an extension to recent approaches. Thorough manual feature engineering allows us to significantly improve current state-of-the-art results. Additionally, we compare applicability of various clustering algorithms to building coreference groups. That is essentially the second part of the approach.