Trading Off Diversity and Quality in Natural Language Generation

TL;DR

This paper frames decoding in natural language generation as a multi-objective optimization problem balancing quality and diversity, providing a comprehensive evaluation of methods and proposing a new algorithm.

Contribution

It introduces a multi-objective framework for decoding evaluation, conducts large-scale comparisons, and proposes the selective sampling algorithm for improved tradeoffs.

Findings

Nucleus sampling outperforms others when quality is prioritized.

All methods perform similarly when diversity is prioritized.

High likelihood sequences can have low quality, known as the 'likelihood trap'.

Abstract

For open-ended language generation tasks such as storytelling and dialogue, choosing the right decoding algorithm is critical to controlling the tradeoff between generation quality and diversity. However, there presently exists no consensus on which decoding procedure is best or even the criteria by which to compare them. We address these issues by casting decoding as a multi-objective optimization problem aiming to simultaneously maximize both response quality and diversity. Our framework enables us to perform the first large-scale evaluation of decoding methods along the entire quality-diversity spectrum. We find that when diversity is a priority, all methods perform similarly, but when quality is viewed as more important, the recently proposed nucleus sampling (Holtzman et al. 2019) outperforms all other evaluated decoding algorithms. Our experiments also confirm the existence of the `likelihood trap', the counter-intuitive observation that high likelihood sequences are often surprisingly low quality. We leverage our findings to create and evaluate an algorithm called \emph{selective sampling} which tractably approximates globally-normalized temperature sampling.