IWSLT 2021

5-6 Aug. 2021 – ACL-IJCNLP 2021

Welcome to IWSLT!

The International Conference on Spoken Language Translation (IWSLT) is the premier annual scientific conference, dedicated to all aspects of spoken language translation. For 18 years running, the conference has published and organized key evaluation campaigns in the field, including the creation of requisite data suites, benchmarks, metrics and key tasks that define progress in the field. This year’s challenges address simultaneous translation, consecutive translation and subtitling tasks, for real-time, low latency as well offline archival purposes and under low-resource, multilingual, and multimodal constraints. Scientific papers and system descriptions documenting the state-of-the art achieved by participating teams and contributing key algorithmic advances are presented at the conference.

IWSLT is the annual meeting for the SIGSLT’s (Special Interest Group on Spoken Language Translation) of ACL, ISCA and ELRA.

IWSLT 2021 will be collocated with ACL-IJCNLP 2021 and will be run as a virtual conference.

Program

Information about keynote speakers and the conference program will be made available nearer to IWSLT 2021.

Shared Tasks

The IWSLT 2021 Evaluation Campaign will host four shared tasks featuring the following focus areas:

Data, metrics, and evaluation will be provided by the conference organizers.

All tasks will follow the same important dates.

Registration

Registration for 2021 Evaluation Campaign is now open: Registration Form.

Please join the IWSLT Evaluation Campaign google group for periodic updates and community questions.

Workshop Session Proposals

Please submit workshop session proposals to IWSLT Organizers. Each proposal should indicate the topic of the workshop, its format and duration (max 2 hours). Please do also provide information about the organizers and potential speakers. The workshop topic should be within the scope of SIGSLT. We are open to any format, e.g. panel, talks, posters, etc. Publication of workshop papers (if any) in the IWSLT proceedings will go through the regular submission channel for scientific papers (see next item).

Paper Submission

Submissions should be scientific papers or shared task system description papers. All papers may consist of 4 to 8 pages of content with additional pages of references, formatted using the ACL-IJCNLP 2021 stylesheets. Upon acceptance, all papers will be given one (1) additional content page to address reviewers’ comments. All papers should be submitted via softconf. Accepted papers will be published in the ACL-IJCNLP 2021 proceedings and will be presented at the IWSLT conference either orally or as a poster.

Note:

  • Scientific papers must be anonymized for submission, while system description papers do not need to be.
  • We accept submissions which have previously been made available as technical reports or preprints, e.g. on arXiv.
  • We will not accept scientific papers for publication that overlap significantly in content or results with papers published elsewhere.
  • Scientific papers which have been, or plan to be, submitted to other venues must be declared as such at the time of submission, and must be withdrawn from the other venues if accepted and published at IWSLT 2021.

Important Dates

Timeline (Updated on May 6, 2021)

Feb 1, 2021 release of shared task training and dev data
Apr 5-23, 2021 evaluation period
Apr 16, 2021 workshop session proposal deadline
Apr 16, 2021 workshop sessions notification
May 13, 2021 paper submission deadline
June 18, 2021 paper notification
Jun 29, 2021 camera ready deadline
Aug 5-6, 2021 IWSLT conference

Note: All deadlines are 11:59PM UTC-12:00.

Organizers

  • Marcello Federico, Amazon, USA (Conference Chair)
  • Alex Waibel, CMU, USA (Conference Chair)
  • Marta R. Costa-jussà, UPC, Spain (Program Chair)
  • Jan Niehues, Maastricht U., Netherlands (Evaluation Chair)
  • Sebastian Stüker, KIT, Germany (Evaluation Chair)
  • Elizabeth Salesky, JHU, USA (Website Chair)

Program Committee

  • Hirofumi Inaguma, Kyoto University, Japan
  • Surafel Melaku Lakew, Amazon AI, USA
  • Carlos Escolano, Universitat Politècnica de Catalunya, Spain
  • Jiatao Gu, Facebook AI Research, USA
  • Changhan Wang, Facebook AI Research, USA
  • Yun Tang, Facebook AI Research, USA
  • Akiko Eriguchi, Microsoft, USA
  • Xian Li, Facebook AI Research, USA
  • Qianqian Dong, Chinese Academy of Sciences, China
  • Yuchen Liu, Princeton, USA
  • Matthias Sperber, Apple, USA
  • Antonio Toral, U. Groningen, Netherlands
  • Thanh-le Ha, Karslruhe Institute of Technology, Germany
  • Krzystof Wolk, Polish‐Japanese Academy of Information Technology, Poland
  • Gerard I. Gallego, Universitat Politècnica de Catalunya, Spain
  • Julian Salazar, Amazon AWS AI, USA
  • Chengyi Wang, Nankai University, China
  • Christian Hardmeier, Uppsala U., Sweden
  • David Vilar, Amazon, Germany
  • Elizabeth Salesky, JHU, USA
  • Evgeny Matusov, AppTek, Germany
  • Juan Pino, Facebook, USA
  • Katsuhito Sudoh, NAIST, Japan
  • Laurent Besacier, IMAG, France
  • Marco Turchi, FBK, Italy
  • Markus Freitag, Google, USA
  • Matteo Negri, FBK, Italy
  • Matthias Huck, LMU, Germany
  • Mattia di Gangi, AppTek, Germany
  • Cuong Hoang, Amazon AI, USA
  • Nguyen Bach, Alibaba, USA
  • Preslav Nakov, QCRI, Qatar
  • Yves Lepage, U. Waseda, Japan
  • Jiajun Zhang,Institute of Automation Chinese Academy of Sciences, China
  • Duygu Ataman, University of Zurich, Switzerland
  • Xutai Ma, Johns Hopkins University, USA
  • Melvin Johnson, Google, USA
  • Cuong Hoang, Amazon, USA
  • José Guilherme Camargo de Souza, University of Trento