Euskorpora is pleased to launch the Basque Speaker Diarization Challenge, an initiative designed to advance research and innovation in speaker diarization for Basque (Euskara), a language that remains significantly underrepresented in current speech technology benchmarks.
Speaker diarization addresses the fundamental question “Who spoke when?” by automatically detecting speaker turns and grouping speech segments belonging to the same individual. Although remarkable progress has been achieved in recent years for high-resource languages, realistic evaluation benchmarks for low-resource languages such as Basque are still scarce.
We invite researchers, academic groups, and industry teams to take part in this challenge and contribute to the development of more robust and inclusive speech technologies.
Challenge Objectives
The Basque Speaker Diarization Challenge aims to:
- Foster research on speaker diarization for Basque.
- Provide a realistic benchmark for low-resource language technologies.
- Encourage the development of systems capable of handling complex broadcast scenarios.
- Promote collaboration within the speech technology community.
Dataset
Participants will have access to a unique corpus of 49 real-world Basque television recordings, comprising approximately 70 hours of audio.
The dataset includes a wide variety of challenging acoustic conditions:
- Television programmes, interviews, debates, and spontaneous conversations.
- Overlapping speech.
- Background noise and music.
- Rapid speaker changes.
- Diverse recording environments and acoustic conditions.
The corpus is divided as follows:
| Split | Recordings | Duration |
|---|---|---|
| Training | 29 | 40 h |
| Development | 8 | 10 h |
| Test | 12 | 20 h |
Evaluation
Submitted systems will be evaluated using internationally recognised speaker diarization metrics:
- Diarization Error Rate (DER) – Official ranking metric.
- Jaccard Error Rate (JER) – Complementary secondary metric.
Important Dates
| Phase | Date |
|---|---|
| Challenge Announcement (Registration Opens) | July 2, 2026 |
| Participation Application Deadline (Registration Form & System Proposal) | July 24, 2026 |
| Release of Training and Development Sets & Evaluation Repository | July 30, 2026 |
| Development Results Submission Deadline | October 9, 2026 |
| Release of Test Set | October 14, 2026 |
| Test Results Submission Deadline | November 4, 2026 |
| System Description Paper Submission Deadline | December 4, 2026 |
| Results Announcement | TBA |
| Workshop Presentation | TBA |
How to Participate
To join the challenge, teams must complete the registration form and submit a short description of their proposed approach.
Detailed participation rules, submission requirements, and evaluation procedures can be found in the challenge-description and the official challenge terms and conditions.
Funding
The Basque Speaker Diarization Challenge includes a total funding budget of €210,000 to support the development of participating systems. Financial support of up to €70,000 per participant will be awarded to selected participants, in accordance with the challenge evaluation and funding criteria described in the official guidelines.
Join the Challenge
The Basque Speaker Diarization Challenge provides a unique opportunity to evaluate cutting-edge diarization systems on realistic Basque broadcast data and contribute to the advancement of speech technologies for low-resource languages.
We look forward to receiving your submissions and building together a stronger research community around Basque speech technology.
