Adapting word order from one language to another is a key problem in cross-lingual structured prediction.
One challenge for evaluating current sequence- or dialogue-level chatbots, such as Empathetic Open-domain Conversation Models, is to determine whether the chatbot performs in an emotionally consistent way.
We investigate the problem of efficiently incorporating high-order features into neural graph-based dependency parsing.
Ranked #12 on Dependency Parsing on Penn Treebank
We describe the graph-based dependency parser in our system (AntNLP) submitted to the CoNLL 2018 UD Shared Task.