Speechocean – update (September 2018):

Speechocean: Global A.I. Data Resource & Service Supplier

About Speechocean

Speechocean is one of the world well-known language related resources & services provider in the fields of Human Computer Interaction and Human Language Technology. At present, we can provide data services with 110+ languages and dialects across the world.

KingLine Data Center---Data Sharing Platform

Kingline Data Center is operated and supervised by Speechocean, which is mainly focused on language resources creating and providing for research and development of human language technology.

These diversified corpora are widely used for the research and development in the fields of Speech Recognition, Speech Synthesis, Natural Language Processing, Machine Translation, Web Search, etc. All corpora are openly accessible for users all over the world, including users from scientific research institutions, enterprises or individuals.

For more detailed information, please visit our website: http://kingline.speechocean.com

New released data:

1. Chinese Mandarin Speech Recognition Corpus

S.N：King-ASR-459

Size: 776 GB

Recording Hours: 2596 Hours

Recording Platform: Desktop

Parameters: 44.1k, 16bit; 2 Channels

Introduction:

The Chinese Mandarin Speech Recognition Corpus was collected in China.

The corpus contains 2,400,698 utterances in total. Each utterance wave was stored in a separate file and uncompressed.

This corpus contains the voices of 2404 different speakers (1184 males, 1220 females) who were balanced distributed in age (18 – 30, 31 – 45, 46 – 60), gender and regional accents. Each speaker was recorded in quiet office and home environments. 

Desktop platform was used for speech collection. A pronunciation lexicon is available with a phonemic transcription in Pinyin. All audio files were manually transcribed and annotated by native transcribers.

2. American English Speech Recognition Corpus

S.N: King-ASR-066

Size: 123 GB

Recording Hours: 1146.6 Hours

Recording Platform: Mobile

Parameters: 16k, 16bit; 1 Channel

Introduction:

This American English Speech Recognition Corpus was collected in USA.

The script contains 829,631 utterances in total. Each utterance wave was stored in a separate file and uncompressed.

This corpus contains the voices of 2,602 different speakers (1,232 males, 1,370 females) who were balanced distributed in age (16 – 30, 31 – 45, >45), gender and regional accents. Each speaker was recorded in quiet office and home environment. 

Mobile platform, i.e. Android, was used for speech collection. A pronunciation lexicon is available with a phonemic transcription in CMU. All audio files were manually transcribed and annotated by native transcribers.

3. Taiwan Chinese Speech Synthesis Corpus - Male

S.N: King-TTS-037

Size: 9.54 GB

Recording Hours: 29.55 Hours

Parameters: 48k, 16bit; 1 Channel

Introduction:

The Taiwan Chinese Speech Synthesis Corpus contains the recordings of 1 male voice talent. He is a broadcaster and he was born and grew up in Taiwan.

The corpus contains 20,736 utterances. It was recorded in a professional studio over one channel of waveform. Speech rate, energy and timbre were strictly controlled during recording process.

Each utterance was carefully proofreaded by linguists and was stored in Windows uncompressed PCM format. Prosody labeling and phone boundary labeling are included. All data were manually checked. 

4. Prounciation Lexicon of Person and Location Name of US English

S.N: King-Lexicon-078

Entries: 359,246

Phoneme Inventory: Computer Readable IPA（It can be converted to the phoneset Sampa, XSampa, and etc., based on demand.)

Stress: Included

Syllable Boundary: Included

Contact Information

Xianfeng Cheng

vice president

Tel: +86-10-62660928;

+86-10-62660053 ext.8080

Mobile: +86-13681432590

Skype: xianfeng.cheng1

Email: chengxianfeng@speechocean.com; cxfxy0cxfxy0@gmail.com

Website: www.speechocean.com