• Establish an approach for effectively working with cultural data across multi-disciplinary teams from diverse disciplines, backgrounds, and skill-sets;
• Shape a flexible and diverse humanities-trained workforce with data science skills, who can manage the social and ethical considerations of cultural data; and
• Develop a shared understanding of the process of cultural data sharing, which honours the labour of content creators as well as those transforming, enriching, and analysing data.

• Using a range of scraping and data wrangling methods, transform existing cultural collections into datasets for immediate use in computational contexts;
• Identify and resolve inconsistencies, redundancies, errors, and gaps in the data;
• Liaise closely with subject matter experts to augment and refine datasets in accordance with industry knowledge, cultural histories, and personal experience;
• Develop conceptual frameworks for the creation of qualitative subsets for use in comparative analysis; and
• Explore new adjacent data sources to complement existing datasets for innovative analytics.

• Investigate and interrogate named entities and ontologies within cultural data collections;
• Engage with existing standards in the digital humanities to develop a prototype information architecture for cultural data;
• Interconnect different cultural data collections without dissolving their distinctive discipline-specific and ontological structures.

• Design innovative models for revealing patterns in Australian arts production, cultural consumption, and tangible and intangible cultural heritage;
• Create new resources for policy formation, cultural production, and research infrastructure development; and
• Extend access to cultural data to facilitate new cross-collection, cross-disciplinary, and cross-institutional research.

National and international efforts to aggregate humanities databases are reshaping concepts of heritage, creativity, and cultural inclusion in the global sphere [Bettivia and Stainforth 2017] [Smithies et al. 2023]. However, as Lisa M. Given, Sarah Polkinghorne and Joann Cattlin point out, “while cultural data initiatives are growing in number, globally, they lack a cohesive, sustainable, and healthy ecosystem to enable collaboration and sharing across related contexts” [Given, Polkinghorne, and Cattlin 2023]. Several international models for the expansion and interoperability of cultural data warrant consideration. The DARIAH-DE repository is a digital archive for the long-term preservation of humanities and cultural research data, aggregated across various services and applications (https://repository.de.dariah.eu/search/). Europeana PRO makes millions of cultural data assets from European galleries, libraries, archives, and museums available for searching and downloading (https://pro.europeana.eu/page/datasets). Similarly, the Digital Repository of Ireland (https://dri.ie/about-us/) was launched in 2015 to facilitate the “preservation, curation, and dissemination of Ireland's humanities, social sciences, and cultural heritage data”. In 2020, the United Kingdom launched Towards a National Collection (TaNC), an £18.9 million, five-year program of investment aimed at “creating a unified virtual national collection” of the UK's museums, libraries, galleries, and archives, “dissolving barriers between different collections”, and “opening UK heritage to the world” [UKRI 2022].

Most of these platforms have elected either to combine data into a single unified search engine or to provide a collection of external links. However, seeking to develop best practices for data sharing and data flows, some international aggregation projects and research groups, such as DHARPA (https://dharpa.org/), have begun to grapple with how to productively interconnect diverse data sources without flattening the distinctiveness of domain-specific collection ontologies. Another dimension to data aggregation and interoperability is the export and generation of cultural heritage datasets, which until recently has been dominated by the Open Archives Initiative Protocol for Metadata Harvesting (https://www.openarchives.org/pmh/), discussed in the section “Program B: Technical Design and Methods” below.

In Australia, the most important digital aggregator of cultural data is Trove, the National Library of Australia (NLA)'s search engine, which over the past several decades has evolved into a powerful platform offering access to the NLA's own collections, as well as other registered collections. While not specifically designed as research cyber-infrastructure, Trove has nevertheless attracted a wide local and international user base, ranging from professional researchers to the general public [Stainforth 2019]. Under its “People and Organisations” portal (https://trove.nla.gov.au/help/categories/people-and-organisations-category), Trove currently links to several cultural collections associated with the ACD-Engine, and NLA software developers as well as independent digital humanities experts have produced a range of tools that enable the searching and downloading of curated research datasets.[1] Historically, however, it has suffered from inadequate resourcing to maintain and renovate core infrastructure and develop research capability [Jones and Verhoeven 2022].[2]

Another important project has been the Humanities Networked Infrastructure (HuNI, https://huni.net.au/), a platform developed as part of the Australian government's National e-Research Collaboration Tools and Researchers programme. The design framework of HuNI recognises humanities data as consisting of “the various annotations, tags, links, associations, ratings, reviews, and comments produced during the humanities research process, together with the semantic ‘entities’ to which these annotations refer: concepts, persons, places, and events” [Burrows 2011]. HuNI's data-centred workflow, in a similar way to Trove, focuses on a “discovery environment” that enables browsing and searching across disparate data sources, though it does not house the data itself [Burrows 2013] [Burrows and Verhoeven 2016]. Australia also has a major centralised data repository, the Australian Data Archive (ADA). The ADA provides a central searchable catalogue of more than 5,000 social science datasets. It holds data from surveys, opinion polls, and censuses from Australia and the Asia-Pacific region (https://ada.edu.au/about-ada/). However, the ADA hosts very little material on arts and culture, and its uncurated, user-submitted datasets are not necessarily immediately suitable for computational analysis. A more encouraging development in the Australian cultural data landscape has been the formation of an Indigenous Data Network in 2018 (https://mspgh.unimelb.edu.au/centres-institutes/onemda/research-group/indigenous-data-network), and, more recently, national investment in a HASS and Indigenous Research Data Commons within the Australian Research Data Commons (ARDC). These projects justly champion the importance of Aboriginal and Torres Straits Islander custodianship of rich cultural traditions as they are expressed in data and housed in cultural collections.

In Australia, as much as in Europe, North America, and the United Kingdom, the demand for quality data and online access to data-rich sources of information has grown exponentially, with more stakeholders entering the field each year. As such, the cultural data landscape in Australia is broad, yet fragmented, and has been supported for decades by fixed-term national investment grants made to individual humanities, arts, and social sciences (HASS)-based digital tools, platforms, working groups, and labs (such as the ARDC's HASS Community Data Lab,: https://ardc.edu.au/project/hass-community-data-lab/). As Mike Jones and Alana Piper point out, the “funding models set up for traditional projects normally rely on them having a clear end date for financial support, after which digital resource may not even be accessible, let alone updated” [Jones and Piper 2023]. Moreover, many individual research collections and digital databases have been built using bespoke technical approaches (often outsourced to private companies), which have led to institution-specific issues relating to design, access, maintenance, and use of data. Inevitably, the diversity of cultural data collections reflects different histories of data collection, management, curation, and maintenance, and can lead to intergenerational and sometimes competitive misalignments of the potentiality in cultural data. Similarly, and in spite of national and international calls for open data policies (see e.g., https://inke.ca/), the economic, cultural, and social data managed by governments is often limited by idiosyncratic data structures, short-term political support, or survey parameters (e.g., https://www.arts.gov.au/cultural-data-online).

Nonetheless, the appetite for accessing, connecting, and telling stories with Australian cultural data in the creative industries continues apace, as evidenced by a reinvigorated national cultural policy, REVIVE, announced in 2023 [Australian Government 2023]. Researchers across the humanities are also increasingly interested in the research possibilities to be realised by using structured cultural data and large digital collections. However, as Ahnert et al. rightly point out in their analysis of collaborative historical work using big data, relatively little research considers “the practical steps in getting hold of such data, and the restrictions with which it may come” [Ahnert et al. 2023].

The interdisciplinary Melbourne team, in consultation with project partners, undertook the day-to-day work of the ACD-Engine project as outlined in the project workflow below. The team comprised the following six part-time posts, totalling approximately three full-time positions.

In additional, we benefited from the expertise of students affiliated to, but not funded by the project:

Cultural researchers from our project partners were involved in the ACD-Engine's reflexive exploration, transformation, and analysis of the cultural data extracted from their databases. Each chief investigator was tasked to the project at either half a day or one day per week, with research assistants on short-term part-time contracts. They provided five primary sources of cultural data across five different artistic fields, many of which developed from ARC investment in research and/or database construction, as follows:

Each database presented distinctive opportunities to interrogate prior assumptions about data entry, coding protocols, and relationality. Taken together, they provided a manageable volume of data for an integrated information architecture to emerge. This under-the-hood approach, secured by the generous commitment of partners to the exploration of their existing databases, facilitated an iterative knowledge-sharing approach between data engineers, subject domain experts, and digital humanities scholars.

Our partner institutions also included teams at Swinburne University's Data Co-Op (one chief investigator), specialising in government data analysis, and the Time-Layered Cultural Map of Australia (TLC Map) (https://tlcmap.org/, see [Arthur et al. 2020], one chief investigator), which provided a platform for geospatial analysis and detailed interrogation of demographic contexts. Researchers from the Social Change Impact Enabling Platform at RMIT University and the King's Digital Lab and Department of Digital Humanities at King's College London (one chief investigator each, along with supporting local staff), provided understanding of and alignment with national and global infrastructure priorities, as well as connections to user groups in Australia and the UK. The ACD-Engine was also supported by an international advisory board comprised of researchers and infrastructure specialists from the University of Melbourne, Flinders University, the National University of Singapore, the University of Ghent, and the Victorian government's Centre for Data Insights, which met approximately three times a year across the project to provide strategic advice and oversight.

We constructed the ACDEA based on the following principles [Liu, McKay, and Buchanan 2023]:

While remaining open to the inclusion of diverse or novel attributes, this model provided us with a view of the range and complexity of data from cultural databases across multiple disciplines, without diminishing the significance of each entity in disciplinary histories. Each of the terms we use for these entities, we argue, has a firm basis in existing cultural research methodologies and an intuitive, difficult-to-dispute accessibility. We arrived at these terms following extensive debates about the meaning of various concepts (for the results of one discussion, see [Cutter, Fensham, and Sumner 2023]). Simultaneously, this data ontology facilitates the potential future integration of datasets from other sources or search engines of specific interest to a research project, whether they might be from urban design, music, or the art market. Each entity in ACDEA contains three essential identifying metadata:

To facilitate use of ACDEA, each entity has been described in a data dictionary [Zheng and Munoz 2023b] and detailed visualisation [Zheng and Munoz 2023c][9] with a simplified version provided in the table below:

Some of the attributes listed above represent discrete attributes, while others (such as “career” under “person”) provide an entry point to further nested data attributes. In addition, the bottom-up construction of the ACDEA ensures that all relevant attributes from the originating data sources are included (even though some attributes do not map across all datasets) without losing sight of larger questions of scale, sustainability, and interoperability with other national and international collections.

The results are showcased in a number of tangible, open-access outputs, including but not limited to:

In the process, we undertook investigations into each partner database, leading to a number of new insights and directions for future research. Each data partner explains some of their key findings below:

As this showcase highlights, with the capability of the ACD-Engine's robust datasets and outputs now readily available, members of the project team have begun to explore a range of novel use-cases for data analysis, particularly where data interrogation might be placed alongside scholarly and topical debates, or interfaced with other smaller datasets that invite new understandings of digital cultural heritage. Further applications might include:

We further anticipate that the datasets will now be used in data analytics pedagogy in the digital humanities, information science, and digital collections. Data scientists and graduate researchers could also experiment with the data coding and models in their own research.