%Aigaion2 BibTeX export from Idiap Publications
%Wednesday 16 September 2026 06:13:35 PM

@INPROCEEDINGS{Baneras-Roux_INTERSPEECH2026_2026,
                      author = {Ba{\~{n}}eras-Roux, Thibault and Burdisso, Sergio and Villatoro-Tello, Esa{\'{u}} and Sanchez-Cortes, Dairazalia and Liu, Shiran and Baroudi, Severin and Kumar, Shashi and Watawana, Hasindri Sankalpana and E, Manjunath K and Hacioğlu, Kadri and Motlicek, Petr and Stolcke, Andreas},
                    keywords = {domain adaptation, low-resource, speech recognition},
                    projects = {Idiap, UNIPHORE, ELOQUENCE},
         mainresearchprogram = {AI for Everyone},
                       title = {Closing the Speech-Text Gap with Limited Audio for Effective Domain Adaptation in LLM-Based ASR},
                   booktitle = {INTERSPEECH 2026},
                        year = {2026},
                    abstract = {Conventional end-to-end automatic speech recognition (ASR) systems rely on paired speech–text data for domain adaptation. Recent LLM-based ASR architectures connect a speech encoder to a large language model via a projection module, enabling adaptation with text-only data. However, this introduces a modality gap, as the LLM is not exposed to the noisy representations produced by the speech projector. We investigate whether small amounts of speech can mitigate this mismatch. We compare three strategies: text-only adaptation, paired speech–text adaptation, and mixed batching (MB), which combines both. Experiments in in-domain and out-of-domain settings show that even limited speech consistently improves performance. Notably, MB using only 10\% of the target-domain (less than 4 hours) speech achieves word error rates comparable to, or better than, conventional ASR fine-tuning with the full dataset, indicating that small amounts of speech provide a strong modality-alignment signal.},
                         pdf = {https://publications.idiap.ch/attachments/papers/2026/Baneras-Roux_INTERSPEECH2026_2026.pdf}
}