Neural-Davidsonian Semantic Proto-role Labeling

TL;DR

This paper introduces Neural-Davidsonian, a bidirectional LSTM-based model for semantic proto-role labeling that achieves state-of-the-art results and efficiently shares parameters across attributes for improved learning.

Contribution

The paper proposes a novel Neural-Davidsonian model that enhances SPRL by representing predicate-argument pairs and sharing parameters across attributes for better learning efficiency.

Findings

Achieved state-of-the-art SPRL performance.

Model naturally shares parameters between attributes.

Enables learning new attribute types with limited supervision.

Abstract

We present a model for semantic proto-role labeling (SPRL) using an adapted bidirectional LSTM encoding strategy that we call "Neural-Davidsonian": predicate-argument structure is represented as pairs of hidden states corresponding to predicate and argument head tokens of the input sequence. We demonstrate: (1) state-of-the-art results in SPRL, and (2) that our network naturally shares parameters between attributes, allowing for learning new attribute types with limited added supervision.