National Speech Corpus

The National Speech Corpus (NSC) is the first large-scale Singapore English corpus spearheaded by the Info-communications and Media Development Authority (IMDA) of Singapore. It aims to become an important source of open speech data for automatic speech recognition (ASR) research and speech-related applications.

There is a growing trend for people to use voice to interact with services, be it at home, at work, or in public spaces. Supporting Speech Technologies can be inaccurate at recognising and transcribing locally accented English. To solve this technology gap, IMDA introduced the National Speech Corpus (NSC).

The NSC improves speech engines’ accuracy of recognition and transcription for locally accented English. The NSC is also able to contribute to speech synthesis technology where an AI voice can be produced that is more familiar to Singaporeans, with local terms pronounced more accurately.

As speech technology improves and with speech engines tuned to the Singaporean English accent, this will enable Singapore to keep pace with current and future advancements for speech interfaces.

For example, the Automatic Speech Recognition (ASR) may be used by telco call centres to transcribe calls for auditing and sentiment analysis purposes, chatbots can go beyond text and can accurately support our accent while replying in a familiar local accent with accurate pronunciations of street names and food.

Click here to download the Corpus.

1. What are the cost associated to using the NSC?

The NSC is made available via the Singapore Open Data Licence.

2. Do I need to have a Dropbox account to download the NSC?

Yes, users who are interested to download the NSC will require a Dropbox account. The shared folder will be shared to users via email.

3. How big is the NSC?

Currently, the NSC is approximately 1.2 TB in size.

4. Will there be future updates to the National Speech Corpus (NSC)? 

There are currently no planned future updates to the baseline corpora. The last update occurred in July 2021, and another 3 parts were added to the corpora.

5. How do I obtain the latest datasets (all 6 parts), I downloaded the corpora when there were only 3 parts.

Simply re-register and you should receive a link to all 6 parts.

6. Is there a demo of the transcription or speech engine?

The speech corpus is a dataset of audio and transcripts, and not a speech engine. Do contact us if you wish to find industry contacts who have trained their speech engine with the NSC. 

For further enquiries on the National Speech Corpus, please contact nsc@imda.gov.sg.