Augmenting BERT-style Models with Predictive Coding to Improve Discourse-level Representations

TL;DR

This paper introduces a method to enhance BERT-style models with predictive coding, enabling better discourse-level understanding and improving performance on discourse-related tasks.

Contribution

It proposes a novel augmentation using predictive coding to improve discourse-level representations in BERT-style models.

Findings

Improves performance on 6 out of 11 discourse-related tasks

Enhances the ability to predict future sentences

Excels in discourse relationship detection

Abstract

Current language models are usually trained using a self-supervised scheme, where the main focus is learning representations at the word or sentence level. However, there has been limited progress in generating useful discourse-level representations. In this work, we propose to use ideas from predictive coding theory to augment BERT-style language models with a mechanism that allows them to learn suitable discourse-level representations. As a result, our proposed approach is able to predict future sentences using explicit top-down connections that operate at the intermediate layers of the network. By experimenting with benchmarks designed to evaluate discourse-related knowledge using pre-trained sentence representations, we demonstrate that our approach improves performance in 6 out of 11 tasks by excelling in discourse relationship detection.