Curricular Transfer Learning for Sentence Encoded Tasks

TL;DR

This paper introduces a curriculum-based transfer learning approach for NLP tasks, improving model adaptation across different data distributions, especially in conversational environments.

Contribution

It proposes a novel sequence of pre-training steps guided by data hacking and grammar analysis to enhance transfer learning in NLP.

Findings

Significant improvement on MultiWoZ task

Effective adaptation across distribution shifts

Outperforms existing pre-training methods

Abstract

Fine-tuning language models in a downstream task is the standard approach for many state-of-the-art methodologies in the field of NLP. However, when the distribution between the source task and target task drifts, \textit{e.g.}, conversational environments, these gains tend to be diminished. This article proposes a sequence of pre-training steps (a curriculum) guided by "data hacking" and grammar analysis that allows further gradual adaptation between pre-training distributions. In our experiments, we acquire a considerable improvement from our method compared to other known pre-training approaches for the MultiWoZ task.