Mandarin Chinese speech synthesis data
About this dataset
1. Product Overview
The Mandarin Chinese speech synthesis database contains a total of 654 hours and 657,000 entries. The recorded texts cover 15 domains, including smart home, autonomous driving, and industrial production. A total of 1,928 speakers from various accent regions across China participated in the recording process. All recordings were conducted in a quiet indoor environment using high-fidelity microphones (44.1 kHz, 16-bit). The audio data for TTS0006 is sourced from this database. Professional phonetic annotators have transcribed the data into pinyin with tone markings, and rigorous quality checks ensure a text accuracy rate exceeding 97%.
2. Recording Corpus
Corpus Pool Construction
Corpus Pool Contents
Considering the applications of speech recognition in smart homes, autonomous vehicles, industrial manufacturing, and other fields, the corpus was curated across 15 domains, each collected with three distinct content types.
Content Type One · Content Type Two · ·
Serial Number · Domain · Serial Number · Domain
1 · Home Appliance Names and Control Commands · 1 · Home Appliances
2 · POI (Geographic Information) · 2 · POI
3 · Music-related Commands · 3 · Music Commands
4 · Digital Commands · 4 · Digital Commands
5 · TV and Movie Titles · 5 · TV Titles
6 · Finance · 6 · Financial Policies
7 · Technology · 7 · Real Estate
8 · Sports · 8 · International Technology
9 · Entertainment · 9 · Domestic Technology
10 · Current Affairs News · 10 · Sports
11 · English Pronunciation · 11 · Western Entertainment
· · 12 · Hong Kong/Taiwan Entertainment
· · 13 · Current Affairs News
· · 14 · General News
· · 15 · English Pronunciation
Corpus Processing
- De-identification: Removing politically sensitive, personally private, or pornographic/violent content.
- Removal of symbols such as <, >, [ , ] , ～ , / , \ , =.
- Exclusion of materials containing languages other than Chinese and English.
- Standardization of formatting.
3. Speaker Information
Basic Information Records
Speaker information includes task ID, gender, age group, and place of origin.
Task ID · Age Group · Gender · Accent Region
SSB0001 · B · Female · Northern Region
Each speaker receives one task ID, corresponding to one recording assignment. Each speaker may participate only once. Age groups are defined as follows: A (under 18), B (18–25), C (26–45), D (45+). Gender: Male—male; Female—female. Accent regions are categorized based on the speaker’s native language area: Northern, Southern, and Other.
Speaker Demographic Characteristics
Gender Ratio
The database comprises 1,928 speakers, including 828 males and 1,100 females.
Gender · Male · Female · Total
Proportion · 43% · 57% · 100%
4. Data Recording Environment
Recording Environment
A quiet indoor setting, free from obvious human voices or other noise, with no echoes. Speakers read aloud their assigned texts at normal speaking speed.
Recording Equipment
The recording equipment consists of high-fidelity microphones. The database uses a 44.1 kHz, 16-bit mono format.
Recording Method
Speakers maintain a distance of 20 cm from the microphone, reading their texts at normal volume and pace. Android and iOS smartphones are positioned 20 cm away from the microphone respectively.
5. Annotation and Transcription Guidelines
Text transcription is based on the spoken audio, aiming to align written content with the actual pronunciation. Key principles include: 1) Transcriptions must exactly match the heard speech—no extra words, omissions, or errors. 2) Numbers should be written in Chinese characters, e.g., “一二三,” rather than “123.” Pay attention to distinguishing between “一” and “幺,” and “二” and “两.” 3) English words appearing within sentences should be transcribed according to pronunciation—for example, “thank you.” Letters pronounced individually should be rendered in uppercase followed by spaces, such as “N B A” or “U F O.” 4) Symbols appearing in sentences should be transcribed as spoken by the speaker—for instance, “三W点百度点com.” Unpronounced symbols should be omitted. Brand names and proper nouns should follow standard conventions, such as “QQ空间,” “iPhone,” or “喜马拉雅.” 5) Annotation completeness must mirror actual pronunciation—no deletions allowed. Pinyin transcription is based on the actual pronunciation heard, following these guidelines: 1) Numbers 1, 2, 3, and 4 represent specific pinyin tones, while 5 denotes neutral tone. For example: SSB00610358.wav becomes “dao4时shi2我wo3会hui4联lian2系xi4你ni3”; SSB00610502.wav becomes “天tian1赐ci4良liang2缘yuan2家jia1具ju4商shang1场chang32.” 2) Two-tone and three-tone variations require careful handling. When two consecutive third tones occur, the first tone often lowers to second tone—for example: “你好吗ni2 hao3 ma5,” “洗澡xi2 zao3,” “缓缓huan2 huan3.” However, not all tones necessarily shift, so transcription must adhere strictly to the recorded audio rather than personal pronunciation habits. 3) “一” and “不” undergo tone changes: “一” retains its original first tone when spoken alone; “不” maintains its fourth tone unless pronounced in non-four-tonal contexts. In cases where “一” precedes a fourth tone, it shifts to second tone—for example: “一yi2半ban4,” “一yi2道dao4,” “一yi2律lv4”; similarly, “不bu2必bi4,” “不bu2但dan4,” “不bu2错cuo4.” When “一” appears before a non-four tone, it assumes the fourth tone—for example: “一yi4般ban1,” “一yi4些xie1,” “一yi4朝zhao1”; when “一” precedes a second tone, it takes on that second tone—for example: “一yi4齐qi2,” “一yi4旁pang2,” “一yi4同tong2”; and when “一” comes before a third tone, it adopts that third tone—for example: “一yi4举ju3,” “一yi4手shou3,” “一yi4体ti3.”
6. Product Directory Structure
Directory Structure
The directory structure organizes the data systematically:
└─DOC（Textual Documentation）
││├─wav_list.txt（Audio File List）
││├─content.txt（Speech Text Content）
││├─spk_info.xlsx（Speaker Information）
│
└─SPEECHDATA（Data Folder）
│├─SSB0001（Speaker Folder）
││ ││ SSB00010001.wav（Audio File）
Naming Conventions
CORPUS/USAGE/SPEAKER_NUM/SPEECH_IDe.g./SPEECHDATA/SSB0001/SSB00010001.wav
Directory Name · Content · Notes
CORPUS · · Speech Database Identifier
USAGE · SPEECHDATA · Folder Name
SPEAKER_NUM · SSB0001 · Speaker Folder
SPEECH_ID · SSB00010001.wav · WAV File
Data Sample Download
Samples
- Audio sample 1
- Audio sample 2
- Audio sample 3
- Audio sample 4