EMNLP Findings2023
Simultaneous Machine Translation with Tailored Reference
Abstract
Simultaneous machine translation (SiMT) generates translation while reading the whole source sentence. However, existing SiMT models are typically trained using the same reference disregarding the varying amounts of available source information at different latency. Training the model with ground-truth at low latency may introduce forced anticipations, whereas utilizing reference consistent with the source word order at high latency results in performance degradation. Consequently, it is crucial to train the SiMT model with appropriate reference that avoids forced anticipations during training while maintaining high quality. In this paper, we propose a novel method that provides tailored reference for the SiMT models trained at different latency by rephrasing the ground-truth. Specifically, we introduce the tailor, induced by reinforcement learning, to modify ground-truth to the tailored reference. The SiMT model is trained with the tailored reference and jointly optimized with the tailor to enhance performance. Importantly, our method is applicable to a wide range of current SiMT approaches. Experiments on three translation tasks demonstrate that our method achieves state-of-the-art performance in both fixed and adaptive policies 1 .
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
9 top-tier papers have cited it since. Know what they found before your reviewers do.
Lune can draw on
the full paper · the top-tier papers that cite it · related work
Free to start. No credit card required.
Lineage
- Unified Segment-to-Segment Framework for Simultaneous Sequence Generation cites Monotonic Multihead Attention
- Unified Segment-to-Segment Framework for Simultaneous Sequence Generation cites Future-Guided Incremental Transformer for Simultaneous Translation
- Unified Segment-to-Segment Framework for Simultaneous Sequence Generation cites Modeling Dual Read/Write Paths for Simultaneous Machine Translation
- Future-Guided Incremental Transformer for Simultaneous Translation cites Monotonic Multihead Attention
- Decoder-only Streaming Transformer for Simultaneous Translation cites Unified Segment-to-Segment Framework for Simultaneous Sequence Generation
- Decoder-only Streaming Transformer for Simultaneous Translation cites Modeling Dual Read/Write Paths for Simultaneous Machine Translation
- Large Language Models Are Read/Write Policy-Makers for Simultaneous Generation cites Monotonic Multihead Attention
- Large Language Models Are Read/Write Policy-Makers for Simultaneous Generation cites Decoder-only Streaming Transformer for Simultaneous Translation
- Large Language Models Are Read/Write Policy-Makers for Simultaneous Generation cites Modeling Dual Read/Write Paths for Simultaneous Machine Translation
- Learning Adaptive Segmentation Policy for Simultaneous Translation cites Monotonic Multihead Attention
- Modeling Dual Read/Write Paths for Simultaneous Machine Translation cites Monotonic Multihead Attention
- Modeling Dual Read/Write Paths for Simultaneous Machine Translation cites Future-Guided Incremental Transformer for Simultaneous Translation
- Modeling Dual Read/Write Paths for Simultaneous Machine Translation cites Learning Adaptive Segmentation Policy for Simultaneous Translation
Cited by top-tier papers9
- Unified Segment-to-Segment Framework for Simultaneous Sequence GenerationShaolei Zhang, Yang FengNeurIPS 2023 · 9 citations
- FastLongSpeech: Enhancing Large Speech-Language Models for Efficient Long-Speech ProcessingShoutao Guo, Shaolei Zhang, Qingkai Fang, Zhengrui Ma et al.NeurIPS 2025 · 4 citations
- Simultaneous Interpretation Corpus Construction by Large Language Models in Distant Language PairYusuke Sakai, Mana Makinae, Hidetaka Kamigaito, Taro WatanabeEMNLP 2024 · 3 citations
- Decoder-only Streaming Transformer for Simultaneous TranslationShoutao Guo, Shaolei Zhang, Yang FengACL 2024 · 3 citations
- Large Language Models Are Read/Write Policy-Makers for Simultaneous GenerationShoutao Guo, Shaolei Zhang, Zhengrui Ma, Yang FengAAAI 2025 · 3 citations
Builds on15
- Monotonic Multihead AttentionXutai Ma, Juan Miguel Pino, James Cross, Liezl Puzon et al.ICLR 2020 · 148 citations
- Directed Acyclic Transformer for Non-Autoregressive Machine TranslationFei Huang, Hao Zhou, Yang Liu, Hang Li et al.ICML 2022 · 82 citations
- Future-Guided Incremental Transformer for Simultaneous TranslationShaolei Zhang, Yang Feng, Liangyou LiAAAI 2021 · 44 citations
- Learning Adaptive Segmentation Policy for Simultaneous TranslationRuiqing Zhang, Chuanqiang Zhang, Zhongjun He, Hua Wu et al.EMNLP 2020 · 41 citations
- Modeling Dual Read/Write Paths for Simultaneous Machine TranslationShaolei Zhang, Yang FengACL 2022 · 27 citations
Related papers
- Universal Simultaneous Machine Translation with Mixture-of-Experts Wait-k PolicyShaolei Zhang, Yang FengEMNLP 2021 · 19 citations
- Learning Optimal Policy for Simultaneous Machine Translation via Binary SearchShoutao Guo, Shaolei Zhang, Yang FengACL 2023 · 9 citations
- LLMs Can Achieve High-quality Simultaneous Machine Translation as Efficiently as OfflineBiao Fu, Minpeng Liao, Kai Fan, Chengxi Li et al.ACL Findings 2025 · 3 citations
- Context Consistency between Training and Inference in Simultaneous Machine TranslationMeizhi Zhong, Lemao Liu, Kehai Chen, Mingming Yang et al.ACL 2024
- Gaussian Multi-head Attention for Simultaneous Machine TranslationShaolei Zhang, Yang FengACL Findings 2022 · 15 citations
