Sichuan Dialect Speech Database

1. Product Overview

The Sichuan Dialect Speech Database contains a total of 688 hours of audio. The data types include read speech and conversational speech, with 341 hours of read speech and 347 hours of conversational speech. The recording language is the Sichuan dialect, and the recordings were made in Sichuan, China. A total of 1,729 speakers were invited to record using their mobile phones in quiet environments. This database has undergone professional proofreading, achieving a text accuracy rate of over 95%, making it suitable for research in areas such as speaker recognition and speech recognition.

2. Recording Corpus

Recording Text Results

In the read-speech portion of this database, each speaker recorded 470 sentences, covering wake-up words, command words, free-form text, and more. The specific distribution is as follows:

Domain · Example Sentences · ID Numbers

Wake-Up Words · Xiao Ai Tongxue, Xiao Du Xiao Du · 0001–0021

POI · Navigate to Science and Technology Square · 0022–0044

Home Automation Control · Please help me turn on Video Two · 0045–0051

Phonetic Spelling · Q M P E · 0052–0054

Numbers · Eight hundred sixty zero nine thousand five hundred seventy-seven · 0055–0074

Q&A · What movies did Chen Shisan star in? · 0075–0089

Music · Sing a song by Wang Sulong · 0090–0119

TV & Movies · The Female Kingdom (Chinese version) · 0120–0139

Radio · I want to tune AM to 915.5 MHz · 0140–0144

Entertainment · Ouyang Nana later worked in mainland China · 0145–0160

Finance · Year-on-year growth of 1.9% · 0161–0240

Technology · Total shipments fell short of Apple Watch’s first-day pre-order volume · 0241–0290

Sports · Going forward, we will base ourselves on North马 · 0291–0340

Current Affairs News · A notice was posted on the glass door · 0341–0390

General · For you, tonight must be very long · 0391–0470

Total · · 470

During conversational recording, speakers freely improvised based on given topics. The topic categories are roughly as follows, with corresponding percentages:

E-commerce & Retail · 18%

Finance · 9%

Education · 12%

Technology · 9%

Life Services · 11%

Automotive Sales · 9%

Real Estate · 10%

Logistics · 6%

Other · 17%

Total · 100%

Corpus Pool Processing

• De-identification processing: Removing politically sensitive content, personal privacy information, pornography, violence, and other inappropriate material.
• Removing symbols such as <, >, [, ], ~, /, \, =, except for numeric strings.
• Deleting content containing languages other than Chinese and English.
• Standardizing formatting.

3. Speaker Information

Gender Ratio

The database includes a total of 1,729 speakers—725 males and 1,004 females.

Gender · Male · Female · Total

Percentage · 41.9% · 58.1% · 100%

Age Distribution

The age distribution of speakers is as follows: A (18–25 years old); B (26–40 years old); C (41–50 years old); D (51+ years old). The percentage of speakers in each age group is shown below:

A · 18–25 years old · 454 · 26.3%

B · 26–40 years old · 1,006 · 58.2%

C · 41–50 years old · 244 · 14.1%

D · 51+ years old · 25 · 1.4%

Total · · 100% ·

4. Data Recording Environment

Recording Environment

A quiet indoor setting, free from echoes and other ambient noise.Recording Equipment

Common smartphones available on the market, such as Apple, Huawei, Xiaomi, OPPO, etc. The data storage format is 16 kHz, 16-bit WAV files.Recording Method

For the read-speech portion, speakers recorded in quiet indoor settings using a smartphone app, reading texts at normal speaking speed. For conversational speech, speakers recorded via phone calls in quiet indoor environments.

5. Speech Data Transcription

Transcribers convert audio into written text, striving to ensure that the transcribed content accurately reflects the spoken audio. General guidelines are as follows: 1) The transcription must exactly match the spoken audio—no extra words, missing words, or errors. 2) Numerals should be transcribed in Chinese characters—for example, “one two three” instead of “123.” Pay special attention to distinguishing between “one” and “yao,” and “two” and “liang.” 3) If the audio contains English pronunciation, it should be transcribed into corresponding Chinese characters or left in English. Specific cases include: All letters or words in URLs should be capitalized—for example, if the audio says “www,” it should be transcribed as “Three W dot A B C dot com.” English words within the audio should be transcribed entirely in lowercase. English letters within the audio should be transcribed entirely in uppercase. For certain proper nouns written in all caps, maintain capitalization without spaces—for example, CEO, NBA, etc. 4) The completeness of annotations must match the actual pronunciation; no omissions allowed.

6. Data File Directory

Directory Structure

The directory tree structure of the dataset is as follows:

└─dialogue（Conversational Speech Data）
│├─DOC（Text Folder）
│├─wav_list.txt（Audio List）
│├─spkrinfo.xlsx（Speaker Information）
│├─Textgrid（Transcription Text Folder）
│├─PCD_0111.Textgrid（Transcribed Text）
│├─SPEECHDATA（Data Folder）
││ ││ PCD_0111.wav（Audio File）
│
└─reading（Read-Speech Data）
│├─DOC (Text Folder)
│├─content.txt（Transcribed Text）
│├─wav_list.xlsx（Audio List）
│├─spkrinfo.xlsx（Speaker Information）
│├─SPEECHDATA (Data Folder)
│├─PC0322 (Speaker ID)
││ ││ PC0322_0001.wav（Audio File）

命名 Rules

Directory Naming Rules

<DATA_TYPE>/<USAGE>/<SpeakerID>/<wav_ID>e.g.reading/SPEECHDATA/PC0322/PC0322_0001.wav

Directory · Content · Notes

DATA_TYPE · dialogue/reading · Data Category

USAGE · SPEECHDATA · Folder Name

SpeakerID · PC0322 · Speaker ID

wav_ID · PC0322_0001.wav · Audio File

File Naming Rules

<SpeakerID>_<wav_ID>e.g. PC0322_0001.wav

File · Content · Notes

SpeakerID · PC0322 · Speaker ID

<SpeakerID>_<wav_ID> · 0001.wav · Audio ID

Data Sample Download

Click to download a sample dataset