analiza aiqbench
This research demonstrates that model merging offers only modest generalization for code-switching ASR across unseen language pairs, revealing a fundamental scalability bottleneck in current approaches. The findings suggest that relying on synthetic data or pair-specific fine-tuning fails to solve the combinatorial explosion of required language pair support. For developers building multilingual voice interfaces, this implies that achieving truly universal code-switching capabilities will likely require architectural shifts beyond simple domain generalization techniques.
Towards Truly Multilingual ASR: Generalizing Code-Switching ASR to Unseen Language Pairs
arXiv:2606.05846v1 Announce Type: new Abstract: Automatic Speech Recognition (ASR) has become a key technology for human--AI interaction. However, code-switching ASR (CS-ASR) remains particularly challenging due to the severe scarcity of multilingual CS speech resources across diverse language pairs. Existing approaches primarily improve CS-ASR performance through synthetic CS speech generation or pair-specific fine-tuning on limited bilingual datasets. Nevertheless, these approaches face an inherent scalability limitation, as support for CS must be developed separately for language pairs whose number grows combinatorially with the number of supported languages. In this work, we investigate whether CS capabilities learned from a limited set of seen language pairs can generalize to unseen language pairs through model merging and domain generalization methods. Our experiments show that merged bilingual CS-ASR models modestly generalize to unseen language pairs, suggesting limited transfer of bilingual CS capabilities across language pairs.
Abstract:Automatic Speech Recognition (ASR) has become a key technology for human--AI interaction. However, code-switching ASR (CS-ASR) remains particularly challenging due to the severe scarcity of multilingual CS speech resources across diverse language pairs. Existing approaches primarily improve CS-ASR performance through synthetic CS speech generation or pair-specific fine-tuning on limited bilingual datasets. Nevertheless, these approaches face an inherent scalability limitation, as support for CS must be developed separately for language pairs whose number grows combinatorially with the number of supported languages. In this work, we investigate whether CS capabilities learned from a limited set of seen language pairs can generalize to unseen language pairs through model merging and domain generalization methods. Our experiments show that merged bilingual CS-ASR models modestly generalize to unseen language pairs, suggesting limited transfer of bilingual CS capabilities across language pairs.
Computation and Language (cs.CL); Audio and Speech Processing (eess.AS)
arXiv:2606.05846 [cs.CL]
(or arXiv:2606.05846v1 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2606.05846 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 234 cuvinte
