This is a summary of my presentations at the conference ACL + IWSLT 2025:
How “Real” is Your Real-Time Simultaneous Speech-to-Text Translation System?
Sara Papi, Peter Polák, Dominik Macháček, Ondřej Bojar
- Published in TACL
- In-person poster presentation at ACL 2025
- On Monday, July 28, 11:00-12:30 CEST, Hall 4/5, Session 2: IP-Posters
- ACL Anthology
Simultaneous Translation with Offline Speech and LLM Models in CUNI Submission to IWSLT 2025
Dominik Macháček, Peter Polák
- System submission for Simultaneous Speech Translation shared task at IWSLT 2025
- In-person poster presentation at IWSLT 2025
- On Thursday, July 31, 14:00–15:30, Poster Session I
- PDF (pre-print)
- Code SimulStreaming
- Code ASR Latency with Continuous Levenshtein Alignment
IWSLT Demo session and panel: New applications of speech translation
I will present a live interactive demo of simultaneous speech translation using SimulStreaming, with prompt for in-domain terminology and context, during the IWSLT Demo session on Thursday, July 31, 16:30-17:30.
The demo is available to anyone to try by themselves through this post.
-----
My other news:
- I'm going to start a post-doc fellowship project "Live Credible Translation" in University of Edinburgh from January 2026. See slides introducing this project.
- I will be at MT Marathon in Helsinki.
- We're going to present a "Corpus of Cross-lingual Dialogues with Minutes and Detection of Misunderstandings." To be published at TSD 2025, dataset already available on GitHub and Lindat.
- AI Interpreting service to anybody who could find it useful. On events, meetings, etc. It's an exciting tech-transfer/business project.