Universal Decompositional Semantic Parsing

TL;DR

This paper presents a transductive model for parsing natural language into Universal Decompositional Semantics graphs, jointly predicting graph structure and semantic attribute scores, with analysis showing it captures attribute relationships.

Contribution

It introduces a novel transductive parser that jointly learns UDS graph structure and attribute annotations, improving semantic parsing capabilities.

Findings

Transductive parser performs comparably to pipeline models.

Model captures natural relationships between semantic attributes.

Joint learning enhances semantic graph annotation accuracy.

Abstract

We introduce a transductive model for parsing into Universal Decompositional Semantics (UDS) representations, which jointly learns to map natural language utterances into UDS graph structures and annotate the graph with decompositional semantic attribute scores. We also introduce a strong pipeline model for parsing into the UDS graph structure, and show that our transductive parser performs comparably while additionally performing attribute prediction. By analyzing the attribute prediction errors, we find the model captures natural relationships between attribute groups.