Accent-Dependent Speech Database
About this dataset
1. Product Overview
The Accent Speech Corpus contains a total of 1,258 hours of recordings in Chinese. A total of 1,493 speakers from various accent regions across China participated in the recording process. Recordings were made in quiet indoor environments using smartphones, with speakers engaging in natural conversational dialogues. This corpus is suitable for research in speaker recognition, speech recognition, and related fields. The data has been transcribed and annotated by professional linguists and rigorously quality-checked, achieving a text accuracy rate exceeding 96%.
2. Recording Corpus
This corpus consists of natural telephone conversations covering common domains such as e-commerce and education. The specific structure is as follows:
Domain · Content · Percentage (%)
E-commerce & Retail · Operator services, investment promotion, returns/exchanges, sales pitches · 18.1
Education · Course consultations · 17.1
Lifestyle Services · Catering, tourism, entertainment, transportation · 28.3
Automotive Sales · Vehicle buying/selling, maintenance, car insurance · 2.76
Real Estate · Property buying/selling, rentals · 6.8
Government · Online government consultation · 1.31
Logistics · Express delivery tracking · 6.03
Other · Free-form conversation · 19.6
Total · 100 ·
3. Speaker Information
Accent Regions
The speakers in this corpus represent seven major dialect groups: Northern Mandarin, Wu, Gan, Xiang, Min, Yue, and Hakka. The breakdown by region is shown below:
Accent Region · Number of Speakers · Percentage (%)
Northern Mandarin · 218 · 14.6
Wu · 277 · 18.5
Gan · 187 · 12.5
Xiang · 221 · 14.8
Min · 192 · 12.9
Yue · 180 · 12.1
Hakka · 218 · 14.6
Total · 1,493 · 100
Gender Ratio
The corpus includes a total of 1,493 speakers—564 males and 929 females.
Gender · Male · Female · Total
Ratio · 37.8 · 62.2 · 100%
Age Distribution
The speakers range in age from 18 to 60, distributed as follows: Group A (18–25), Group B (26–40), and Group C (41+).
· Age Group · Number of Speakers · Percentage (%)
Group A · 18–25 · 782 · 52.4
Group B · 26–40 · 496 · 33.2
Group C · 41+ · 215 · 14.4
Total · 1,493 · 100 ·
4. Data Recording Environment
Recording Environment
A quiet indoor setting, free from echoes and external noise.
Recording Equipment
Common Android smartphones available on the market, including Huawei, OPPO, vivo, Xiaomi, etc. Data is stored in 8 kHz or 16-bit WAV format.
Recording Method
Speakers record naturally via their phones in a quiet indoor environment.
5. Speech Data Transcription
Transcribers convert audio into written text, striving to ensure consistency between spoken content and written form. General guidelines include: 1) Transcriptions must exactly match the spoken words—no extra or missing words, no errors. 2) Numbers should be spelled out in Chinese characters—for example, “one two three” instead of “123.” Distinguish between “one” and “yao,” and between “two” and “liang.” 3) English pronunciations should be rendered in corresponding Chinese characters or left as English. Specific cases are as follows: All letters or words in URLs should be capitalized—for instance, “www.abc.com” becomes “Three W dot A B C dot com.” English words within the audio should be transcribed entirely in lowercase. English letters appearing in the audio should be capitalized. For certain proper nouns or acronyms, capitalize all letters and add spaces—for example, “C E O,” “C C T V,” etc. 4) Annotation completeness must align with actual pronunciation; no omissions allowed. 5) If dialectal terms appear in sentences, preserve meaning by substituting homophones and marking them with . 6) In cases of overlapping speech between two individuals, only transcribe the primary speaker’s words while adding an tag. Whether the overlap is continuous or intermittent, mark each overlapping word with . Examples: Example 1: Person A says, “The weather today is so hot!” Before finishing, Person B responds, “Mm.” Since “mm” overlaps with “hot” and both voices are similar in volume, transcription becomes: “The weather today is so hot!” Example 2: Person A says, “I think it’s like this, mm.” Person B replies, “I think so too.” Since Person B’s response overlaps with Person A’s “mm” before completion, transcription becomes: “I think it’s like this.” and “I think so too.”
6. Data File Directory
The directory structure of the dataset is as follows:
└─DOC (Text Description Folder)
││├─wav_list.txt (Audio List)
││├─spkrinfo.xlsx (Speaker Information)
│
└─SPEECHDATA (Data Folder)
│├─00001 (Call Number)
│├─00001.wav (Audio File)
│├─00001.TextGrid (Transcription Text)
Data Sample Download
Samples
- Audio sample 1