Speech Language Models in Low-Resource Settings:
Performance, Evaluation, and Bias Analysis

Co-located with LREC 2026 • Full-Day Workshop

Welcome to SPEAKABLE 2026!

A full-day workshop on Speech Language Models in Low-Resource Settings, focusing on Performance, Evaluation, and Bias Analysis.

Submit Your Paper Learn More

Quick Overview

SPEAKABLE 2026 brings together researchers and practitioners working on speech-native language models, with a special focus on low-resource settings. The workshop addresses the persistent constraints on data availability, annotation quality, and computational budget that affect underrepresented languages and speaker communities.

"Build strong models, measure what matters, and make bias analysis routine for speech in the long tail."

Workshop Countdown

Workshop date to be announced

Check the Important Dates page for submission deadlines and workshop schedule.

Invited Speakers

Jordi Luque (Confirmed)
Lead Research Scientist, Telefónica Research
Title of talk: TBD

Three Core Strands

1. Efficient Adaptation

Parameter-efficient methods, multilingual transfer, knowledge distillation, and edge-constrained inference for low-resource speech tasks.

2. Meaningful Evaluation

Moving beyond WER to task-appropriate metrics, calibration analysis, and slice-aware reporting by accent, dialect, and channel.

3. Responsible Practice

Treating bias analysis as routine scientific reporting, with transparent data documentation and privacy guardrails.

Contact

For questions or more information about the workshop, please contact the organizing committee:

Email:
speakable2026@gmail.com

For more information about the organizing team, visit the Organizers & Committee page.