Skip to main navigation menu Skip to main content Skip to site footer

Best-First Beam Search

Abstract

Decoding for many NLP tasks requires an effective heuristic algorithm for approximating exact search since the problem of searching the full output space is often intractable, or impractical in many settings.  The default algorithm for this job is beam search--a pruned version of breadth-first search.  Quite surprisingly, beam search often returns better results than exact inference due to beneficial search bias for NLP tasks.  In this work, we show that the standard implementation of beam search can be made up to 10x faster in practice. Our method assumes that the scoring function is monotonic in the sequence length, which allows us to safely prune hypotheses that cannot be in the final set of hypotheses early on. We devise effective monotonic approximations to popular nonmonontic scoring functions, including length normalization and mutual information decoding.  Lastly, we propose a memory-reduced variant of Best-First Beam Search, which has a similar beneficial search bias in terms of downstream performance, but runs in a fraction of the time.

Article at MIT Press Presented at EMNLP 2020

Author Biography

Clara Meister

PhD student under Ryan Cotterell at ETH Zurich