# Speculative Beam Search for Simultaneous Translation

arXiv: https://arxiv.org/abs/1909.05421 · HTML: https://mlsanity.com/paper/1909.05421/

## Key facts

- Cited by 10 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 12 Sep 2019.

## Built on by (most cited first, 10 of 10)

- [RealTranS: End-to-End Simultaneous Speech Translation with Convolutional Weighted-Shrinking Transformer](https://mlsanity.com/paper/2106.04833/) (2021)
- [The USTC-NELSLIP Systems for Simultaneous Speech Translation Task at IWSLT 2021](https://mlsanity.com/paper/2107.00279/) (2021)
- [Alternate Endings: Improving Prosody for Incremental Neural TTS with Predicted Future Text Input](https://mlsanity.com/paper/2102.09914/) (2021)
- [Simultaneous Multi-Pivot Neural Machine Translation](https://mlsanity.com/paper/2104.07410/) (2021)
- [Simul-LLM: A Framework for Exploring High-Quality Simultaneous Translation with Large Language Models](https://mlsanity.com/paper/2312.04691/) (2023)
- [Using fine-tuning and min lookahead beam search to improve Whisper](https://mlsanity.com/paper/2309.10299/) (2023)
- [Segmentation-Free Streaming Machine Translation](https://mlsanity.com/paper/2309.14823/) (2023)
- [Adversarial Machine Learning in Text Analysis and Generation](https://mlsanity.com/paper/2101.08675/) (2021)
- [Visualization: the missing factor in Simultaneous Speech Translation](https://mlsanity.com/paper/2111.00514/) (2021)
- [Improving Stability in Simultaneous Speech Translation: A Revision-Controllable Decoding Approach](https://mlsanity.com/paper/2310.04399/) (2023)

## Abstract (arXiv)

Beam search is universally used in full-sentence translation but its application to simultaneous translation remains non-trivial, where output words are committed on the fly. In particular, the recently proposed wait-k policy (Ma et al., 2019a) is a simple and effective method that (after an initial wait) commits one output word on receiving each input word, making beam search seemingly impossible. To address this challenge, we propose a speculative beam search algorithm that hallucinates several steps into the future in order to reach a more accurate decision, implicitly benefiting from a target language model. This makes beam search applicable for the first time to the generation of a single word in each step. Experiments over diverse language pairs show large improvements over previous work.