A review of recent work on Simultaneous MT

Glad to announce that we’re finally restarting our weekly lab meetings and reading group after a brief hiatus. Going forward, we’ll have our lab meetings on Wednesdays, 11:30 AM.

For this week’s online lab meeting Aditi will present a review of some of the recent approaches on Simultaneous MT.

Simultaneous Machine Translation defines a task of generating high-quality translation under low latency. Existing SiMT methods are divided into fixed (Wait-k, Multi-Path Wait-k, MoE Wait-k) and adaptive policies (MiLk, MMA etc.). She will be presenting an overview of the various approaches taken towards this task, and specifically present these 5 papers :

1. Ma et al., “Monotonic Multihead Attention”, ICLR 2020
2. Zhang et al., “Wait-info Policy: Balancing Source and Target at Information Level for Simultaneous Machine Translation”, EMNLP 2022
3. Zhang et al., “Modeling Dual Read/Write Paths for Simultaneous Machine Translation”, ACL 2022
4. Alinejad et al., “Translation-based Supervision for Policy Generation in Simultaneous Neural Machine Translation”, EMNLP 2021
5. Deng et al., “Improving Simultaneous Machine Translation with Monolingual Data”, AAAI 2023

Wednesday, 15 March at 11:30.

The zoom link will be posted on zulip.