Reimagining Access: Inclusive Technology Design for Archives and Special Collections
- Sustainable Development
by Nikita Pashenkov
Creating transcriptions for audiovisual content is an excellent way to make materials more accessible. However, creating transcriptions manually is costly and time-consuming, especially for archives that have large collections. For the Reimagining Access project, we piloted Whisper, an Automatic Speech Recognition (ASR) application by OpenAI that uses AI and pulls from thousands of hours of audio on the web for its dataset
Our assessment from the pilot was that Whisper is incredibly accurate, but not without flaws, sometimes adding additional text not spoken to the transcript. We realized that the human element is still crucial for editing and formatting. However, creating a complete transcript with Whisper takes significantly less time than creating one manually. We envision this tool (and other ASRs) to be used in the workflow of processing audiovisual materials, as important as other steps in processing digital and analog materials.
The Reimagining Access GitHub page includes installation requirements and instructions for Whisper.
https://github.com/artcenter-interaction/reimagining-access/blob/main/docs/index.md#whisper-gui