In October 2016 the NSW State Library launched a new platform for delivering digitised audio materials, Amplify.
The online platform enables users to immerse themselves in fascinating moments throughout NSW’s history by viewing photographs, listening to audio files and reading computer-generated transcripts.
Amplify is more than just an online catalogue of library archives. It also captures corrections users make when reading transcripts through an easy-to-use editing interface. The edits are then fed back into the system using machine learning technology – reducing the number of errors in future computer-generated transcripts and enabling more audio collections to be released.
Amplify has broken new ground on many fronts. Borrowing open source code from the New York Public Library, the NSW State Library has been able to publicly release more than 145 historical audio recordings for the first time.
It is also the first time any cultural institution in Australia has used machine learning to enhance access to its audio collections.
Amplify has produced impressive results in a short time. Of the 145 recordings available on the platform in the first six months, 31 individual recordings were 100% corrected by one or more contributors and 74 recordings were 50% or more corrected.
Help improve the State Library's collections, or learn more about the NSW State Library's other digital projects here.
Between October 2016 and June 2017 more than 50,000 corrections have been made to NSW State Library’s recently released audio collections by over 3,000 users from 42 different countries.
March 2018 Update
The Snowy Mountain Bushfires collection of interviews of residents, farmers, children, volunteers and other community members affected by the 2003 fires that blazed for nine weeks has been added to Amplify. This is the sixth oral history collection made available on the Library’s crowdsourced transcription and audio listening platform. We are currently developing enhancements to pilot this with NSW public libraries under a software-as-a-service model.