Tibetan Speech Database

1. Product Overview

The Tibetan speech database comprises a total of 1,204.8 hours of audio. It features a sampling rate of 16,000 Hz and a bit depth of 16 bits. The data has been transcribed and annotated by professional speech annotators and subjected to rigorous quality assurance, achieving a text accuracy rate exceeding 95%.

2. Speech Data Transcription

Transcribers generate textual content based on the audio they hear, striving to ensure that the written text accurately reflects the spoken words. General guidelines are as follows: 1) The transcription must match the audio exactly—no extra words, missing words, or errors. 2) English words should be written in lowercase; phonetic spellings must use uppercase letters with spaces, e.g., “W W W dot Google dot com,” “C E O.” 3) The transcription must fully capture all spoken content without omissions.

3. Directory Structure and Naming Conventions

Directory Structure

Data directory tree structure:

└─DOC (textual documentation)
││├─content.txt (transcription list)
└─WAV (audio folder)
│├─Amdo (region)
││├─T0360987.wav (audio file)

Naming Conventions

CORPUS/USAGE/REGION/SPEECH_IDe.g./WAV/Amdo/T0360987.wav

Directory Name · Content · Notes

CORPUS · · Speech database ID

USAGE · WAV · Audio folder

REGION · Amdo, Kham, Ü-Tsang · Geographic region

SPEECH_ID · T0360987.wav · Audio file

4. Regional Distribution

The database covers Amdo, Kham, and Ü-Tsang regions, with detailed durations and percentage shares shown below:

Region · Duration (hours) · Percentage (%)

Amdo · 483.6 · 40.14

Kham · 209.3 · 17.37

Ü-Tsang · 511.9 · 42.49

Total · 1,204.8 · 100

Sample Data Download

Click to download the sample dataset (data_sample.zip)