Learning to Write with Coherence From Negative Examples

TL;DR

This paper introduces a novel training method called writing relevance (WR) that enhances the coherence of generated text by contrasting positive and negative examples, outperforming existing approaches in human evaluations.

Contribution

The paper presents a new WR training approach that improves coherence in neural text generation by leveraging negative examples, outperforming Unlikelihood training.

Findings

WR training improves coherence in text generation.

Human evaluations favor WR over UL.

WR effectively models coherence by contrasting positive and negative continuations.

Abstract

Coherence is one of the critical factors that determine the quality of writing. We propose writing relevance (WR) training method for neural encoder-decoder natural language generation (NLG) models which improves coherence of the continuation by leveraging negative examples. WR loss regresses the vector representation of the context and generated sentence toward positive continuation by contrasting it with the negatives. We compare our approach with Unlikelihood (UL) training in a text continuation task on commonsense natural language inference (NLI) corpora to show which method better models the coherence by avoiding unlikely continuations. The preference of our approach in human evaluation shows the efficacy of our method in improving coherence.