T. Mizumoto

Dialect-Robust Speech Language Models with Synthetic Pseudo-Dialect Augmentation

P pseudo-dialect speech is synthesized by converting LLM-generated dialect text via a standard-language TTS model, requiring zero real dialect speech, and intermediate standard-text prediction during training is introduced during training, acting as semantic normalization for downstream tasks.

Evaluating the Effect of Question Wording Variations on Answer Consistency in Large Language Models