The models in this demo are built with an approach to parsing as sequence labeling, developed within the FASTPARSE project. The main idea of this approach is that we have developed encodings to represent a full parse trees (including both constituency and dependency trees) for a sentence of length n as a sequence of n labels. This allows us to use any generic, out-of-the-box sequence labeling framework (in our case, we are using NCRF++) to perform full parsing in a fast and accurate way.
More details can be found in Papers section.