Learn and Don't Forget: Adding a New Language to ASR Foundation Models
This addresses the challenge of expanding multilingual ASR systems for low-resource language communities, but the approach is incremental as it builds on existing adaptation techniques.
The paper tackled the problem of adding a new, often low-resource, language to foundation ASR models without degrading performance on existing languages, finding that direct fine-tuning improves new language performance but harms existing ones, while Elastic Weight Consolidation can mitigate this for specific languages.
Foundation ASR models often support many languages, e.g. 100 languages in Whisper. However, there has been limited work on integrating an additional, typically low-resource, language, while maintaining performance on the original language set. Fine-tuning, while simple, may degrade the accuracy of the original set. We compare three approaches that exploit adaptation parameters: soft language code tuning, train only the language code; soft prompt tuning, train prepended tokens; and LoRA where a small set of additional parameters are optimised. Elastic Weight Consolidation (EWC) offers an alternative compromise with the potential to maintain performance in specific target languages. Results show that direct fine-tuning yields the best performance for the new language but degrades existing language capabilities. EWC can address this issue for specific languages. If only adaptation parameters are used, the language capabilities are maintained but at the cost of performance in the new language.