Submissions

Submissions must include both a WER report presented in an ASRU paper and a transcript of the test set. The deadline for submission is 10th July. No new submissions will be accepted after this date, however, authors may make revisions of submitted papers until Friday 17th July as long as the title and author list remain unchanged.

Submitting the paper

Challenge papers should be prepared following the guidelines presented on the ASRU website, http://www.asru2015.org and submitted using ASRU's paper submission system. When submitting a paper there is a drop-down field for "Session Topic" - please choose the CHiME challenge task session. Participants should also email a copy of the paper to chimechallenge@gmail.com. If you revise your paper after the 10th July then please email the revised version on or before the 17th July deadline for revisions.

  • Please check that your paper clearly reports results using the official training and test sets described on the instructions page. However, any additional results are also welcome.
  • Results for both the real and simulated data should be reported but systems will be ranked based on performance on the real test data only.
  • Papers will be reviewed by through the normal ASRU review process.
  • Authors of papers that are rejected will still be able to present their systems during the challenge Special Session.

Submitting the transcripts

Participants must submit the ASR transcripts corresponding to their official results (including transcript for both the real and simulated data) by emailing them to chimechallenge@gmail.com on or before the submission deadline. If your transcripts are revised after the 10th July then please email the revised versions on or before the 17th July deadline for revisions.

If you are using the Kaldi baseline, the transcripts to be submitted are the files called exp/***/best_wer_enhanced.result provided by the scoring functions (local/chime3_calc_wers.sh, local/chime3_calc_wers_smbr.sh).

If you are not using Kaldi, please stick to the transcription format below: 

<utterance id> word_1 word_2, …
For example, 

F01_050C0103_BUS_SIMU CONSTRUCTION EMPLOYMENT FELL FORTY SEVEN THOUSAND AFTER FIFTEEN THOUSAND JOB DECLINE THE MONTH BEFORE
F01_050C0103_CAF_SIMU CONSTRUCTION EMPLOYMENT FELL FORTY SEVEN THOUSAND AFTER A FIFTEEN THOUSAND JOB DECLINE THE MONTH BEFORE
F01_050C0103_PED_SIMU CONSTRUCTION EMPLOYMENT FELL FORTY SEVEN THOUSAND AFTER FIFTEEN THOUSAND JOB DECLINE THE MONTH BEFORE
:
: