Track trials, findings, and clinical evidence as it lands.
Search papers, labs, and topics across Lattice.
Adapting Whisper-small to the Earnings22 benchmark for ASR of financial earnings calls revealed that task complementarity between new and old data actually improved performance on held-out financial speech, negating the need for continual learning regularization. When regularization was applied, a layer-wise mini-KFAC approximation outperformed diagonal EWC in reducing word error rates and stabilizing optimization. These findings underscore the importance of assessing task complementarity before implementing continual learning constraints and suggest mini-KFAC as a superior regularization technique when necessary.
Continual learning regularization can actively harm ASR performance on financial earnings calls due to task complementarity, with mini-KFAC proving more effective than EWC when regularization is unavoidable.
Earnings call transcription presents a realistic continual learning scenario where Automatic Speech Recognition (ASR) systems are updated with new recordings but risk catastrophic forgetting of previously learned financial speech patterns. Adapting Whisper-small from a general-domain speech corpus to the Earnings22 benchmark reveals the opposite: all evaluated methods exhibit improvement on held-out financial speech from the same domain during adaptation rather than forgetting. This task complementarity makes regularization counterproductive. Nevertheless, when regularization is warranted, a layer-wise mini-Kronecker-Factored Approximate Curvature (mini-KFAC) approximation seems to achieve lower word error rates (WERs) and more stable optimization than standard diagonal Elastic Weight Consolidation (EWC). The results suggest that task complementarity should be assessed before applying continual learning constraints, and prefer mini-KFAC EWC over diagonal EWC when needed.