What can we learn from Semantic Tagging?

TL;DR

This paper explores multi-task learning with semantic tagging as an auxiliary task across NLP tasks, demonstrating that selective sharing strategies significantly improve performance and reduce negative transfer.

Contribution

It introduces the 'learning what to share' approach, optimizing task sharing in multi-task learning to enhance NLP task performance.

Findings

Significant improvements in NLP tasks using semantic tagging as auxiliary.

The 'learning what to share' setting yields consistent gains.

Reduced negative transfer with selective sharing strategies.

Abstract

We investigate the effects of multi-task learning using the recently introduced task of semantic tagging. We employ semantic tagging as an auxiliary task for three different NLP tasks: part-of-speech tagging, Universal Dependency parsing, and Natural Language Inference. We compare full neural network sharing, partial neural network sharing, and what we term the learning what to share setting where negative transfer between tasks is less likely. Our findings show considerable improvements for all tasks, particularly in the learning what to share setting, which shows consistent gains across all tasks.