Splitting EUD graphs into trees: A quick and clatty approach
This is an incremental approach for the EUD Shared Task, addressing parsing efficiency and accuracy for computational linguistics researchers.
The authors tackled the problem of parsing Enhanced Universal Dependencies (EUD) graphs by splitting them into trees based on linguistic criteria and combining predictions from a sequence-labelling parser, resulting in relatively poor performance that could be improved with further refinement.
We present the system submission from the FASTPARSE team for the EUD Shared Task at IWPT 2021. We engaged in the task last year by focusing on efficiency. This year we have focused on experimenting with new ideas on a limited time budget. Our system is based on splitting the EUD graph into several trees, based on linguistic criteria. We predict these trees using a sequence-labelling parser and combine them into an EUD graph. The results were relatively poor, although not a total disaster and could probably be improved with some polishing of the system's rough edges.