Members of the Audiovisual Metadata Platform (AMP)<https://go.iu.edu/amppd> project team at Indiana University Libraries (IU), the University of Texas at Austin, AVP, and the New York Public Library (NYPL) are pleased to announce the release of a public white paper presenting the findings of the Audiovisual Metadata Platform Pilot Development (AMPPD) project, which has worked to enable more efficient generation of metadata to support discovery and use of digitized and born-digital audio and moving image collections.
The overarching goal of the AMPPD project, conducted from October 2018 to June 2021, was to develop enough of the AMP system to be able to pilot test it using two audiovisual (AV) collections from IU and a third collection from NYPL. The project team has developed a software system that harnesses the Galaxy workflow engine, originally developed for data processing workflows in computational genomics, to design and execute custom workflows for metadata and feature extraction from AV files. As part of this work, the team also evaluated metadata generation mechanisms (MGMs) in categories such as speech-to-text, video OCR, and named entity recognition, to select open source and commercial cloud solutions for workflow pipelines combining machine learning and human mediation. Based on work and results so far, the project team has concluded that the approach taken in AMP is effective and scalable for generation of metadata for certain types of AV collections, particularly those that involve significant amounts of spoken word content.
This project and the planning project that preceded it in 2017 have been generously supported by funding from the Andrew W. Mellon Foundation, with substantial in-kind staff and computing
contributions from IU.
The report is available for download and viewing at https://hdl.handle.net/2022/26989
Feedback on the report is very much welcomed and may be sent to any AMP project team member.
- Jon Dunn, on behalf of the AMPPD project team
Jon Dunn (he/him)
Assistant Dean for Library Technologies
Indiana University Bloomington Libraries
[log in to unmask]<mailto:[log in to unmask]>