DIarization and Speech Processing for LAnguage understanding in Conversational Environments
Focusing on multilingual medical conversational speech between community health workers and patients
Download the Flyer
The DISPLACE-M challenge aims to advance diarization and speech understanding technologies in real-world healthcare conversations. The dataset includes multilingual, code-mixed interactions between Accredited Social Health Activists (ASHAs) and patients, representing authentic primary healthcare settings.
Three core tracks for 2026
Determine “who spoke when” in code-mixed multilingual medical dialogues with overlapping speech.
Transcribe multilingual, code-mixed healthcare conversations across dialects and noisy environments.
Identify conversation topics (e.g., maternal health, infection)
Produce concise medical dialogue summaries.
Multilingual medical dialogue corpus
~30 hours of real primary-healthcare interactions recorded between community health workers (ASHAs) and patients in local clinics, homes and health camps.
Hindi, English naturally mixed across turns, reflecting true medical multilingual usage.
Single-channel 16 kHz 16-bit WAV recordings, far-field microphones, real acoustic conditions, annotated for speaker segments, language, and overlap.
Evaluation plan and baselines will be provided
Links to datasets, evaluation metrics, and baseline systems will appear here once released.
Download and email the signed Terms & Conditions to displace2026@gmail.com.
After verification, you’ll receive credentials for the development set and baselines.
Meet the team behind DISPLACE-M
Register for the challenge and send the signed agreement to displace2026@gmail.com to receive access credentials.
No, redistribution is prohibited. You may use it for research with proper citation.
No hard constraints on overall compute or number of models or latency constraints.
Yes, you are allowed to use proprietary/open source data for training your models, but it must be properly cited.
Yes, in a prescribed format- will be provided to the registered participants.
That's great! You can follow the Interspeech 2026 paper submission portal here. Remember to select "DISPLACE-M special session” while uploading your paper there.
