autoregressive asr 1catastrophic forgetting 1replay-based editing 1self-supervised correction 1timestamp drift 1
From the 1 of 3 linked papers with an AI index.
3 papers
cs.CL2026
REDDIT: Correcting Model-Generated Timestamp Drift in ASR without Forgetting via Replay-Based Distribution Editing
Cheng-Kang Chou, Ming-To Chuang, Ke-Han Lu +2
The paper investigates drift in model-generated timestamps for autoregressive ASR systems and introduces REDDIT, a replay‑based distribution editing post‑training method that corre…
eess.AS2025
ToxicTone: A Mandarin Audio Dataset Annotated for Toxicity and Toxic Utterance Tonality
Yu-Xiang Luo, Yi-Cheng Lin, Ming-To Chuang +9
Despite extensive research on toxic speech detection in text, a critical gap remains in handling spoken Mandarin audio. The lack of annotated datasets that capture the unique proso…
cs.SD2025
How to Learn a New Language? An Efficient Solution for Self-Supervised Learning Models Unseen Languages Adaption in Low-Resource Scenario
Shih-Heng Wang, Zih-Ching Chen, Jiatong Shi +6
The utilization of speech Self-Supervised Learning (SSL) models achieves impressive performance on Automatic Speech Recognition (ASR). However, in low-resource language ASR, they e…