National Archives seeks AI and ML tech

The National Archives of Australia have released a tender for the procurement of artificial intelligence and machine learning technologies for its archival workflows.

National Archives is the custodian of Australian government records, which in addition to a large physical collection, has a digital collection approaching 10 petabytes of data, comprising a wide range of digital file formats.

As the digital collection is expected to expand considerably over the next few years both through increasing digital transfers and digitisation of the physical collection, National Archives is seeking to conduct and evaluate technology pilots across the areas of transcription, description, access examination and search & discovery.

National Archives are seeking AI/ML capabilities that can transcribe digital records into a text format, so that the transcription can then be stored in a relational database as part of the metadata for the record.

The transcription would then be used in a range of scenarios including enhancing the ability for users to find records containing specific words or phrases; translation of the transcript into foreign languages; and presentation of the transcript to users with accessibility requirements.

National Archives has a set of moderately complex metadata catalogues, stored in relational databases, that enable the management of the digital collection. Metadata records are manually created as items are added to the digital collection, a labour intensive process with a significant backlog of items that are incompletely described.

The agency is therefore seeking solutions that would analyse the content of items in the digital collection and create/update associated metadata records.

When members of the public apply for access to an item in the National Archives collection, the item is examined to determine whether it can be released wholly or partially (with redaction) based on a set of criteria.

This process is also labour intensive with a significant backlog of items to be assessed, so the agency is seeking solutions that would analyse the content of items in the digital collection against a set of criteria based on the Archives Act and highlight for a reviewer the areas of an item that may contain information that falls within one of the exemption categories.

National Archives customers and staff also have a need to find items in the collection.

Existing search capabilities are based on simple keyword matching, only use metadata (not item content) and rely on the user’s understanding of the metadata schema. A search capability with a natural language interface that is able to provide relevant search results to users, based on collection metadata and digital item content, would significantly improve the searchability of the collection, according to the agency.

The tender closes on 27 October 2025 at 5:00 pm (ACT Local Time).