Dependency Graph Parsing as Sequence Labeling

TL;DR

A range of unbounded and bounded linearizations are defined that can be used to cast graph parsing as a tagging task, enlarging the toolbox of problems that can be solved under this paradigm.

Abstract

Various linearizations have been proposed to cast syntactic dependency parsing as sequence labeling. However, these approaches do not support more complex graph-based representations, such as semantic dependencies or enhanced universal dependencies, as they cannot handle reentrancy or cycles. By extending them, we define a range of unbounded and bounded linearizations that can be used to cast graph parsing as a tagging task, enlarging the toolbox of problems that can be solved under this paradigm. Experimental results on semantic dependency and enhanced UD parsing show that with a good choice of encoding, sequence-labeling dependency graph parsers combine high efficiency with accuracies close to the state of the art, in spite of their simplicity.

Figures (5)

• Figure 1: An example of a relaxed $2$-planar dependency graph linearized with our unbounded encodings.
• Figure 2: Bounded encodings for the example of Figure \ref{['fig:example1']}. The relaxed 1-planar subgraphs for the $4k$-bit-encoding are shown with their linearization, added null arcs are drawn with dotted lines, and their associated bits underlined. For $6k$-bit, we use colors to distinguish the subgraph pairs. Note that, in both cases, those arcs that are assigned to different planes w.r.t. the unbounded bracketing encoding (Figure \ref{['fig:example1']}) are marked with $\ast$.
• Figure 3: Pareto front: UF vs. speed. X-axis rescaled, outliers omitted for clarity.
• Figure 4: Prediction steps of our neural parser for the sentence "There were many pioneer PC contributors" (from the DMen treebank) using the 1-planar bracketing encoding.
• Figure 5: Pareto front for the rest of treebanks.