What are the  biggest performance bottlenecks in this model?

This question is tagged as a feature request, since it is an improvement. I see, that the dependency
parser is taking a long time to process a text. What are the greatest performance bottlenecks?