| --- |
| license: cc-by-4.0 |
| --- |
| ### Teochew-MSA（most standard accent，最标准口音）数据集 |
|
|
| 本数据集源自“[风入松文旅]((https://www.douyin.com/user/MS4wLjABAAAAiFta2mV4lAnryJ5bHkCIldv5X6RFYQRltcnUhz3iy_WNJlXg0awlXHS8KYYY-9Kz?from_tab_name=main))”的677条视频，经数据管线处理后，提取出时长适宜的音频片段，并配有由OCR识别生成的初步字幕，总时长约29小时。 |
|
|
| 发音人使用的是标准的潮州府城音，口齿清晰、发音规范，且能熟练运用文白异读，尤其是发音保留了清晰的闭口韵（即国际音标中的 -p、-m 韵尾，该发音特征在现代潮汕话人群中已大量流失），整体音频质量上乘，可以认为是**最标准的潮州话**。 |
|
|
| #### 潜在局限性： |
|
|
| 由于原视频带有背景音乐（BGM），部分中国传统乐器的声音难以被完全消除，可能会对纯净度产生轻微影响。 |
|
|
| #### 元数据（Metadata）： |
|
|
| * 总时长：约 29 小时 |
| |
| * 样本数量：31,298 个音频片段 |
| |
| * 平均时长：3.3 秒/样本 |
| |
| * 音频参数：采样率 44,100 Hz，单声道（Mono）WAV 格式，16-bit PCM 编码 |
|
|
|
|
|
|
| ### Teochew-MSA (Most Standard Accent) Dataset |
|
|
| This dataset is derived from 677 videos produced by "Fengrusong Culture and Tourism". After being processed through a data pipeline, it yields audio clips of appropriate lengths paired with rough subtitles generated via OCR, totaling approximately 29 hours. |
|
|
| The speaker uses the standard Chaozhou prefectural accent (Fucheng accent) with clear articulation and standard pronunciation, demonstrating high proficiency in literary and colloquial readings. Notably, the speaker retains clear closed-syllable codas (i.e., the -p and -m endings in the International Phonetic Alphabet, a phonetic feature that has been lost among many modern Teochew speakers). |
|
|
| Overall, the audio quality is exceptionally high, and it can be considered **the most standard Teochew**. |
|
|
|
|
|
|
| #### Potential Limitations: |
|
|
| Because the original videos contain background music (BGM), the sounds of certain traditional Chinese instruments are difficult to remove completely, which may slightly affect the audio purity. |
|
|
| #### Metadata: |
|
|
| * Total Duration: ~29 hours |
| * Sample Count: 31,298 audio clips |
| * Average Length: 3.3 seconds per sample |
| * Audio Specs: 44,100 Hz sample rate, Mono WAV format, 16-bit PCM encoding |
|
|
