Workshop Program
May 11, 2026
Room: Cabrera 4-5
Times are local (Palma).
09:00 – 09:15
Introduction and general remarks
Chair: Nina Hosseini-Kivanani — RTL & University of Luxembourg
09:15 – 10:15
Keynote: Jordi Luque
Toward Trustworthy SpeechLLMs for Low-Resource Multilingual ASR: Evidence from ELOQUENCE on Adaptation, Bias, and Federated Learning
Chair: Alessio Brutti — FBK, Italy
Abstract, bio, and affiliation · Slides (PDF)
10:15 – 10:30
Remote posters
Chair: Nina Hosseini-Kivanani — RTL & University of Luxembourg
5 min each
Adapting Foundational ASR Models to Efik: An Empirical Study of an Extremely Low-Resource Tonal Language
PAREDA: A Multi-Accent Speech Dataset of Natural Language Processing Research Discussions
HARNESS: Lightweight Distilled Arabic Speech Foundation Models
10:30 – 12:00
Coffee break and poster session
Chair: Alessio Brutti — FBK, Italy
📍 MENORCA Hall
Say Again? The Limits of Whisper with Conversation. A Case Study on the KIParla Corpus.
Not All Polar Questions Are the Same: ASR, Humans, and Russian
Quantizing Whisper: How Design Choices Affect ASR Performance
“OK Aura, Be Fair with Me”: Demographics-Agnostic Training for Bias Mitigation in Wake-up Word Detection
Scalable Expansion of Multilingual Speech LLMs for ASR: A Continual Learning Approach
Responsible Benchmarking of Fairness for Automatic Speech Recognition
Addressing Accent Disparities in Automatic Speech Recognition: A Comparative Study of Single and Two-Step Adaptation
Investigating Speaker Pronunciation Variability in Speech Embeddings: Speaker and L1 Effects on French as a Second Language
What LID Systems Say About Dialectal Variation. The Case of Yiddish, Quechua and Mande
HARNESS: Lightweight Distilled Arabic Speech Foundation Models
When Does OmniASR Fail? A Fine-Grained Human Evaluation on Saudi Arabic Dialects
12:00 – 13:00
Spotlight papers
Chair: Marco Matassoni — FBK, Italy
SpeechLM for Automatic Speech Recognition in Low-resource Languages
Improving Low-resource ASR Using Bilingual Fine-tuning with Language Identification: A Cross-linguistic Evaluation
13:00 – 14:00
Lunch break
14:00 – 16:00
Architectures and learning methods
Chair: Christoph Schommer — University of Luxembourg
15 min + 5 QA
Leveraging Speech Models for Audio-based Lexical Retrieval in Dictionaries: The Case of the Teochew Language
Stage-Aware Cross-Lingual Transfer for Faroese ASR: When and Which Languages Matter
Doing More with Less: Determining Optimal Pre-training Model for Irish Automatic Speech Recognition through Multi-step Fine-tuning
Blank-Aware Decoding for Transcript-Free Phoneme Alignment in Low-Resource Languages and Dialects
On the Role of Encoder Depth: Pruning Whisper and LoRA Fine-Tuning in SLAM-ASR
TaLK-Corpus: A Regionally Diverse Evaluation Set for Sri Lankan Tamil Speech
16:00 – 16:30
Coffee break
16:30 – 17:00
Best paper and closing remarks
Chair: Nina Hosseini-Kivanani — RTL & University of Luxembourg
Presentation slides & posters
Presentation slides
Due to the size of the conference rooms, it is recommended to use 36 pt fonts for the presentation slides.
Keynote slides
- Jordi Luque — Toward Trustworthy SpeechLLMs for Low-Resource Multilingual ASR: Evidence from ELOQUENCE on Adaptation, Bias, and Federated Learning (PDF)
Posters
The size of poster holders is 90 cm × 150 cm and the format is vertical (portrait). The poster boards cannot accommodate landscape posters. You can print your poster in portrait A0 (84.1 × 118.9 cm).