A Simple, Fast Diverse Decoding Algorithm for Neural Generation

TL;DR

This paper introduces a simple, fast decoding algorithm that enhances diversity in neural generation tasks like dialogue, summarization, and translation, with an adaptive RL-based variation further improving performance.

Contribution

It presents a novel diversity-promoting decoding algorithm with an RL-based adaptive variant, improving neural generation quality across multiple tasks.

Findings

Diverse decoding improves task performance, especially with reranking.

The RL-based adaptive method further boosts results.

The approach is effective in dialogue, summarization, and translation.

Abstract

In this paper, we propose a simple, fast decoding algorithm that fosters diversity in neural generation. The algorithm modifies the standard beam search algorithm by adding an inter-sibling ranking penalty, favoring choosing hypotheses from diverse parents. We evaluate the proposed model on the tasks of dialogue response generation, abstractive summarization and machine translation. We find that diverse decoding helps across all tasks, especially those for which reranking is needed. We further propose a variation that is capable of automatically adjusting its diversity decoding rates for different inputs using reinforcement learning (RL). We observe a further performance boost from this RL technique. This paper includes material from the unpublished script "Mutual Information and Diverse Decoding Improve Neural Machine Translation" (Li and Jurafsky, 2016).