speech-corpus

Lipi-Ghor introduces an 885-hour, diarized, multi-domain Bengali speech corpus built from YouTube videos via a replicable pipeline, along with benchmarks of four ASR models and a human-verified gold evaluation split released under CC-BY-4.0.

该论文提出一种噪声鲁棒的强制对齐方法，用上采样WLM特征替代MFCC的GMM-HMM对齐器，并显式建模填充词和非词汇语音事件，在Buckeye和FluencyBank上将词边界误差降低19%和17%，大幅减少灾难性错配。

BanglaKontho presents a 20-hour single-speaker Bangla TTS corpus from professional audiobooks, along with an open-source text normalizer and preprocessing pipeline, to address the gap in long-form prosody for low-resource Bangla speech synthesis.