National Archives of Australia Enterprise Information Manager Project

The Client: National Archives of Australia

The National Archives of Australia (NAA) can best be described as the memory of our nation – collecting and preserving Australian Government records that reflect our history and identity.

Their collection traces events and decisions that have shaped the nation and the lives of Australians. As well as preserving our history, the National Archives plays a key role in helping to ensure the Australian Government and its departments are effective and accountable to the people.

The Challenge: Automated extraction of metadata and classification of records

NAA staff were manually appraising and disposing of records, a very time consuming, expensive and inconsistent process. NAA required a robust system for automatically recommending disposal actions using document content and metadata.   The image below shows the complexity of the manual process.

Semantic Sciences responded with an ambitious proposal to develop a process and technology that aimed to close the wide gap between general disposal authorities such as AFDA Express (a streamlined version of the Administrative Functions Disposal Authority AFDA) and the automatic document sentencing. The proposed enterprise information management tool would support automatic extraction of relevant metadata and auto-classification of records for document sentencing and disposal.

The core challenges Semantic Sciences addressed were:
1. The need to create metadata for documents by automatic inference;
2. Creating auto-classifiers that require little training data, are accurate and whose operation is understandable by people;
3. The need for a simple methodology and technology for creating “gold standard” sentencing decisions for sets of documents which could be used to train and assess the performance of the structured auto-classifier.

The Solution: Enterprise Information Management Tool

Semantic Sciences provided a solution that split the overall complex problem into several straightforward activities, with simple interactive tools for each one. The approach was fast, testable and accurate.

Developed by Semantic Sciences, the Sintelix text and data analytics platform delivered a lot of the capability required by NAA. To support iterative development of automatic sentencing, Semantic Sciences supplemented Sintelix’s existing capabilities with three innovations:

1. A specialised taxonomy viewer/editor for layered taxonomies.

2. A highly productive search-based process for rapidly creating “gold standard” allocations of documents to nodes on the taxonomy tree; and

3. A development system for “structured classification” which coherently combines a network of simple classifiers to determine document sentencing.

The image below shows the auto-classifier tool developed for NAA.