cantonese-radio

## Cantonese Radio Pseudo-Transcription Dataset
- Contains 14k hours of audio sourced from Archive.org
- Columns
- `order_index`: Represents the order of the audio compared to those from the same `filename`
- `link`: Link of the original full audio
- `transcript_whisper`: Transcribed using `Scrya/whisper-large-v2-cantonese` with `alvanlii/whisper-small-cantonese` for speculative decoding
- `transcript_sensevoice`: Transcribed using `FunAudioLLM/SenseVoiceSmall`
- used [OpenCC](https://github.com/BYVoid/OpenCC) to convert to traditional chinese
- isolated event tags to `event_sensevoice`
- isolated emotion tags to `emotion_sensevoice`
- `snr`: Signal-to-noise ratio, extracted from `ylacombe/brouhaha-best`
- `c50`: Speech clarity, extracted from `ylacombe/brouhaha-best`
- `emotion`: Emotion, extracted from `emotion2vec/emotion2vec_plus_large`
- Note that `id` does not reflect the ordering of the audio within the same video
- Processing
- The full audio is split using [WhisperX](https://github.com/m-bain/whisperX), using `Scrya/whisper-large-v2-cantonese`
- it is split in <30s chunks and according to speakers
- No filtering or additional audio processing was done for this dataset
- Filtering is recommended for your own use

# 粤语广播伪转录数据集
- 包含源自互联网档案馆（Archive.org）的14000小时音频素材
- 字段说明
- `order_index`：代表同一份`filename`对应音频的排列顺序
- `link`：原始完整音频的链接
- `transcript_whisper`：采用`Scrya/whisper-large-v2-cantonese`模型，并结合`alvanlii/whisper-small-cantonese`进行推测性解码生成的转录文本
- `transcript_sensevoice`：采用`FunAudioLLM/SenseVoiceSmall`模型生成的转录文本，具体处理流程包括：
- 使用[OpenCC](https://github.com/BYVoid/OpenCC)将文本转换为繁体中文
- 提取事件标签并存入`event_sensevoice`字段
- 提取情感标签并存入`emotion_sensevoice`字段
- `snr`：信噪比，从`ylacombe/brouhaha-best`中提取得到
- `c50`：语音清晰度，从`ylacombe/brouhaha-best`中提取得到
- `emotion`：情感标签，从`emotion2vec/emotion2vec_plus_large`中提取得到
- 注意事项：`id`字段无法反映同视频内音频的实际排序
- 数据处理流程
- 完整音频通过[WhisperX](https://github.com/m-bain/whisperX)工具，结合`Scrya/whisper-large-v2-cantonese`模型进行切割：切割为时长小于30秒的片段，并按说话人进行区分
- 本数据集未进行任何过滤或额外音频处理操作，建议使用者根据自身需求自行完成过滤处理

搜集汇总

背景与挑战

背景概述

该数据集是一个粤语广播伪转录数据集，包含约14,000小时从Archive.org获取的音频，总大小约589.10GB。数据集提供了使用Whisper和SenseVoice模型生成的转录文本，并包含信噪比、语音清晰度和情感分析等辅助信息，音频已按说话人和30秒片段进行分割处理。