CoolFace
Datasetpublic

lianghsun/zh-taiwan

Dataset Card for zh-taiwan zh-taiwan 是一個繁體中文之語音資料集,總計 2,740 筆音頻(train 2,698 / val 14 / test 28),音訊取樣率為 16 kHz。每筆資料包含音頻、音頻長度、繁體中文文本與對應之正規化(簡體)文本,適用於繁體中文之語音合成(TTS)或語音辨識(ASR)模型訓練與評測。 本資料集原始來源為 ivanzhu109/zh-taiwan,本 repository 僅作為鏡像與格式整理之版本,原始著作權歸原作者所有。 Dataset Details Dataset Description 本資料集提供 train / val / test 三個子集,每筆資料包含下列欄位: audio:16 kHz WAV 音頻; text:繁體中文文本,其中英文詞彙以全大寫形式保留(如 FIREFOXONANDROID、GOOGLE); normalized_text:對應之簡體中文正規化文本,保留相同英文大寫形式;… See the full description on the dataset page: https://huggingface.co/datasets/lianghsun/zh-taiwan.

sourceHugging Faceapache-2.0updated 5mo agoView on Hugging Face
1likes23downloads
Dataset Card

Dataset Card for zh-taiwan

<!-- Provide a quick summary of the dataset. --> zh-taiwan 是一個繁體中文之語音資料集,總計 2,740 筆音頻(train 2,698 / val 14 / test 28),音訊取樣率為 16 kHz。每筆資料包含音頻、音頻長度、繁體中文文本與對應之正規化(簡體)文本,適用於繁體中文之語音合成(TTS)或語音辨識(ASR)模型訓練與評測。

本資料集原始來源為 ivanzhu109/zh-taiwan,本 repository 僅作為鏡像與格式整理之版本,原始著作權歸原作者所有。

Dataset Details

Dataset Description

<!-- Provide a longer summary of what this dataset is. --> 本資料集提供 train / val / test 三個子集,每筆資料包含下列欄位:

  • audio:16 kHz WAV 音頻;
  • text:繁體中文文本,其中英文詞彙以全大寫形式保留(如 FIREFOXONANDROIDGOOGLE);
  • normalized_text:對應之簡體中文正規化文本,保留相同英文大寫形式;
  • duration:音頻長度(秒)。

同時保留繁體與簡體版本之文本,便於繁中語音模型於訓練時使用繁體輸出,也保留與原始簡體文本之可比對性。

  • Curated by: Liang Hsun Huang
  • Language(s) (NLP): Traditional Chinese / Simplified Chinese
  • License: Apache 2.0

Dataset Sources

<!-- Provide the basic links for the dataset. -->

Uses

<!-- Address questions around how the dataset is intended to be used. -->

Direct Use

<!-- This section describes suitable use cases for the dataset. -->

本資料集主要設計用於:

  • 繁體中文語音合成(TTS)模型之訓練與評測;
  • 繁體中文自動語音辨識(ASR)模型之訓練或評測;
  • 研究中文文本正規化與語音對齊之處理流程;
  • 作為小型繁中語音任務之原型驗證素材。

Out-of-Scope Use

<!-- This section addresses misuse, malicious use, and uses that the dataset will not work well for. --> 本資料集不適用於下列用途:

  • 作為大規模語音預訓練之主要語料,因資料量僅約 2,700 筆。
  • 作為口音或方言研究之權威素材,本資料集未附帶說話者人口統計資訊。
  • 涉及語音生物辨識或聲紋追蹤之應用,本資料集不提供說話者身份標註。
  • 作為台語(閩南語)或客語之訓練資料,本資料集僅涵蓋華語。

Dataset Structure

<!-- This section provides a description of the dataset fields, and additional information about the dataset structure such as criteria used to create the splits, relationships between data points, etc. -->

python
{
  "duration": 7.24,
  "text": "這可能是第一個FIREFOXONANDROID的版本CREDIT。",
  "normalized_text": "这可能是第一个FIREFOXONANDROID的版本CREDIT。",
  "audio": {"bytes": <WAV bytes>, "path": "..."}
}
欄位型別說明
durationfloat64音頻長度(秒)
textstring繁體中文文本(英文詞彙以全大寫保留)
normalized_textstring正規化之簡體中文文本
audioAudio音頻檔(16 kHz WAV,內嵌於 parquet)
切分筆數資料大小
train2,698~515 MB
val14~2.5 MB
test28~5 MB

Dataset Creation

Curation Rationale

<!-- Motivation for the creation of this dataset. -->

繁體中文之公開語音資料集長期不足,且既有資料多以簡體中文為主,對繁中 TTS 與 ASR 模型之訓練造成瓶頸。本資料集提供繁體中文版本之文本與對應音頻,作為繁中語音任務之小型基礎素材。

Source Data

<!-- This section describes the source data (e.g. news text and headlines, social media posts, translated sentences, ...). -->

Data Collection and Processing

<!-- This section describes the data collection and processing process such as data selection criteria, filtering and normalization methods, tools and libraries used, etc. -->

原始資料來自 ivanzhu109/zh-taiwan。音頻為 16 kHz WAV 格式,文本經文字正規化(英文詞彙大寫、標點規整)並同時保留繁體與簡體兩個版本。資料依原始 repository 之切分方式分為 train / val / test 三個子集。本 repository 僅為鏡像與 parquet 格式整理。

Who are the source data producers?

<!-- This section describes the people or systems who originally created the data. It should also include self-reported demographic or identity information for the source data creators if this information is available. -->

原始音頻與文本由 ivanzhu109/zh-taiwan 之作者 ivanzhu109 蒐集並釋出,本資料集未附帶說話者身份、性別、年齡等人口統計資訊。

Annotations

<!-- If the dataset contains annotations which are not part of the initial data collection, use this section to describe them. -->

Annotation process

文本正規化(normalized_text)為自動處理之結果,將英文詞彙統一為全大寫並規整標點符號。繁體文本為對應之繁中版本。

Who are the annotators?

不適用(自動處理)。

Personal and Sensitive Information

<!-- State whether the dataset contains data that might be considered personal, sensitive, or private (e.g., data that reveals addresses, uniquely identifiable names or aliases, racial or ethnic origins, sexual orientations, religious beliefs, political opinions, financial or health data, etc.). If efforts were made to anonymize the data, describe the anonymization process. --> 本資料集不包含說話者身份資訊。音頻內容為一般新聞或公開發表之文本,未涉及個人隱私。

Bias, Risks, and Limitations

<!-- This section is meant to convey both technical and sociotechnical limitations. -->

  • 資料量小(總計 2,740 筆),不足以單獨訓練大型語音模型。
  • 文本中英文詞彙以全大寫形式保留,對某些 tokenizer 或 grapheme-to-phoneme(G2P)處理流程可能需要額外適配。
  • 缺少說話者 ID、性別、年齡、腔調等人口統計標籤。
  • 16 kHz 之取樣率低於現代 TTS 模型常用之 22/24 kHz,用於生成任務時可能影響音質。
  • 文本內容之領域分佈與時效性未明確標註,模型可能受限於特定主題範圍。

Recommendations

<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->

建議使用者:

  • 將本資料集作為小型驗證素材,搭配其他繁中語音資料集(如 TAT、Common Voice zh-TW)進行大規模訓練;
  • 若需 22 kHz 或更高取樣率之音頻,應評估上採樣品質之影響;
  • 使用前檢視 textnormalized_text 之大寫規則是否符合下游流程之需求。

Citation

<!-- If there is a paper or blog post introducing the dataset, the APA and Bibtex information for that should go in this section. -->

bibtex
@misc{zh-taiwan,
  title        = {zh-taiwan: Traditional Chinese Speech Dataset},
  author       = {Liang Hsun Huang},
  year         = {2026},
  howpublished = {\url{https://huggingface.co/datasets/lianghsun/zh-taiwan}},
  note        = {Traditional Chinese speech dataset with 2,740 audio clips (train/val/test) at 16 kHz.}
}

Dataset Card Authors

Liang Hsun Huang

Dataset Card Contact

Liang Hsun Huang