Generating event descriptions under syntactic and semantic constraints

TL;DR

This paper evaluates methods for generating event descriptions that meet specific syntactic and semantic constraints, comparing manual, corpus-based, and language model approaches for their naturalness, typicality, and distinctiveness.

Contribution

It provides a comprehensive comparison of three generation methods under constraints, highlighting their strengths and limitations for semantic annotation tasks.

Findings

Manual generation yields the most natural and distinctive descriptions.

Automated methods produce sufficiently high-quality descriptions for downstream tasks.

All methods reliably generate descriptions meeting the specified constraints.

Abstract

With the goal of supporting scalable lexical semantic annotation, analysis, and theorizing, we conduct a comprehensive evaluation of different methods for generating event descriptions under both syntactic constraints -- e.g. desired clause structure -- and semantic constraints -- e.g. desired verb sense. We compare three different methods -- (i) manual generation by experts; (ii) sampling from a corpus annotated for syntactic and semantic information; and (iii) sampling from a language model (LM) conditioned on syntactic and semantic information -- along three dimensions of the generated event descriptions: (a) naturalness, (b) typicality, and (c) distinctiveness. We find that all methods reliably produce natural, typical, and distinctive event descriptions, but that manual generation continues to produce event descriptions that are more natural, typical, and distinctive than the automated generation methods. We conclude that the automated methods we consider produce event descriptions of sufficient quality for use in downstream annotation and analysis insofar as the methods used for this annotation and analysis are robust to a small amount of degradation in the resulting event descriptions.