A Simple, Fast Diverse Decoding Algorithm for Neural Generation
This work addresses the need for more diverse and less repetitive outputs in neural generation for applications like dialogue and translation, though it appears incremental as it builds on existing beam search methods.
The paper tackles the problem of generating diverse outputs in neural generation tasks by proposing a simple, fast decoding algorithm that modifies beam search with an inter-sibling ranking penalty, and it shows improvements across dialogue response generation, abstractive summarization, and machine translation, with a reinforcement learning variation providing further performance boosts.
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 This paper includes material from the unpublished script "Mutual Information and Diverse Decoding Improve Neural Machine Translation" (Li and Jurafsky, 2016).