gMan: Creating General-Purpose Virtual Environments for (Digital) Archival Research

gMan: Creating General-Purpose Virtual Environments for (Digital) Archival Research
Blanke, Tobias, Centre for e-Research, King’s College London, tobias.blanke@kcl.ac.uk
Connor, Richard, University of Strathclyde, richard.connor@cis.strath.ac.uk
Hedges, Mark, Centre for e-Research, King’s College London, mark.hedges@kcl.ac.uk
Kristel, Conny, Netherlands Institute for War Documentation (NIOD), c.kristel@niod.knaw.nl
Priddy, Mike, Centre for e-Research, King’s College London, priddy@mac.com
Simenoni, Fabio, University of Strathclyde, fabio.simeoni@cis.strath.ac.uk
This paper will present a critical analysis of our attempts to build Virtual Research Environments (VREs) for everyday Humanities research tasks using digital archives. Numerous specialised VREs have been developed for addressing particular tasks in various humanities disciplines. The Silchester VREs addressed data integration in archaeological excavations, the SDM VRE developed services for sharing and annotating manuscripts, while TEXTvre is concerned with TEI-based resource creation. Building on these experiences, gMan addressed the issue of moving beyond support for specific, focused tasks, and instead building services to enable more general-purpose humanities research activities, such as integrating and organising the heterogeneous and often unstructured digital resources, and support for ‘active reading’ processesBrockman et al. 2001. Scholarly work in the humanities and the evolving information environment. Washington, DC. through advanced discovery facilities. Such services regularly top the list of humanities user requirementsBenardou et al. 2009. Understanding the Information Requirements of Arts and Humanities Scholarship. International Journal of Digital Curation.. This paper describes work to this end, firstly by the DARIAH project, and subsequently consolidated by the gMan project, funded by JISC’s VRE Rapid Innovation programme.

These experiments were based on use cases identified by the earlier LaQuAT (Linking and Querying Ancient Texts) projectJackson et al. 2009. Building bridges between islands of data—an investigation into distributed data management in the humanities. Proceedings of the Fifth IEEE International Conference on e-Science. Washington, DC., which investigated how to integrate scattered, heterogeneous and autonomous data resources relating to ancient texts, mainly databases but also XML corpora. LaQuAT attempted to solve these issues by offering an integration framework based on the OGSA-DAI grid middleware, which provided an integrated interface to the various data resources that followed a relational database model. However, this approach had certain limitations for our purposes, as such models are optimised for dealing with datacentric resources - that is, resources consisting primarily of structured data such as numbers, dates or very short text fields - rather than text-centric resources containing significant quantities of unstructured text. The approach worked well where the structural context of the information was clear and the query aimed at exact matches. More commonly, however, humanities researchers work with text-centric resources, perhaps enhanced with XML mark-up to capture document structure and additional metadataNentwich, M. 2003. Cyberscience. research in the age of the internet. Vienna., and they look for resources for further investigation based on looser criteria of relevance, e.g. by searching for all Roman legal texts in one resource containing information on punishments that are also mentioned in papyri from another resource.

These conclusions were further elaborated in the use cases that were developed from them, which are the main drivers for the work described here. Complementing this is a body of methodological investigation concerning scholars and their use of sources, particularly their use of data and archives. Before describing our current work, we will survey briefly these investigations.

The difference in scholarly practices between the sciences and the mainstream humanities is highlighted in a studyPalmer et al. 2009. Scholarly information practices in the online environment. that investigated the types of information sources used in different humanities disciplines, based on results from the US Research Libraries Group reports. Structured data is relatively little used, except in some areas of historical research, and data as it is traditionally understood in the sciences, e.g. the results of measurements, even less so. It is true that the study is partly outdated, and that data in the traditional sense is increasingly important in the humanities, particularly in linguistics and archaeology where scientific techniques have been widely adopted. Nevertheless, it is clear that in general humanities research relies not on measurements as a source of authority, but rather on the provenance of sources and peer-assessment, and that what data repositories are for the sciences, archives are for the humanitiesDuff et al. 2004 Historians’ use of archival sources: Promises and pitfalls of the digital age. The Public Historian.. Archival records are primary sources about the past and may take many forms, including government papers, financial documents, photographs, sound recordings, etc. All this information is unstructured in nature.

Thus, our work is driven partly by the requirements fromDuff et al. 2004 Historians’ use of archival sources: Promises and pitfalls of the digital age. The Public Historian., interpreted so as to relate to methods of research in archives. Retrieval is to happen in real time, and traditional finding aids are to be complemented by more sophisticated retrieval mechanisms, including the ability to create relevance indexes on unstructured resources, as well as the ability to combine resources in new ways. In particular, we aimed to implement the personal copy of a finding aid that is often quoted as an important prerequisite for specialised research in archives.

Our work investigated how (digital) archival content can be delivered to humanities researchers more effectively, independently of the location and implementation of that content, and with special facilities provided for customising the retrieval, management and manipulation of the content. We investigated how the UK and European research infrastructure (RI) can be exploited to support data-driven, collaborative research in the humanities by using the gCube environmentCandela et al. 2009. On-demand Virtual Research Environments and the Changing Roles of Librarians. Library Hi Tech., which was developed by the EU-funded D4Science project. gCube allows virtual research communities to deploy VREs on demand by making use of the shared resources of the European RI, and provides services that match closely the sort of information organisation and retrieval activities that we identified as being typical in humanities research.

D4Science provides an easy way of scavenging online data resources. It has a consistent mechanism to import data for rich user interaction within the deployed VREs. Its data resource staging framework, based on a well-defined workflow of data analysis, data modelling and data generation, is one of the key innovations of D4Science. The analysis and modelling phases define how data collections are loaded into gCube compound objects using its simple but powerful data model. In the data generation phase, descriptive metadata and provenance information are added.

Using the gCube data staging framework, the following datasets were brought together in our experiments:

The Heidelberger Gesamtverzeichnis (HGV) der griechischen Papyrusurkunden Aegyptens, a collection of metadata records for 65,000 Greek papyri from Egypt.
Projet Volterra, a database of Roman legal texts, currently in the low tens of thousands but very much in progress, stored in a series of themed tables in Microsoft Access.
The Inscriptions of Aphrodisias, a corpus of about 2,000 ancient Greek
These datasets were the same as those used in the LaQuAT project and thus and allow a critical comparison of results. They overlap in terms of time, places and people – specifically looking at the first five centuries or so of the Roman Empire – although their contents are otherwise quite different. The provision of an environment for working with this data in an integrated form would be highly fruitful for the researcher.

The presentation will describe the use cases that we used for evaluating gCube. Our approach was to break down the scenarios identified in interviews at KCL and within DARIAH into a number of common, atomic actions. Specific instances of these actions can be combined to model a variety of "real" research scenarios, for example the ability to assemble heterogeneous resources (or parts of resources) into a virtual collection, to share this virtual collection within a specific community and to search across a virtual collection, where specific search parameters (such as the importance of specific locations) can be set according to preference. Specific communities also require specific search services such as geo-referenced and date-range searches. Finally, the researcher wants to share links between research objects and annotations (including related documents publications) in her community.

In our experiments, we confirmed that most of these use cases could be supported by the features already provided by the core D4Science systems. For the Digital Humanities 2011 presentation, we will address our subsequent activities: the analysis that we carried out to identify gaps in the existing service provision; some results that demonstrate a clear distinction between the viewpoints of humanities and science research, in respect of such features as image search; our move to develop gMan as a production service for humanities researchers; and the recently-funded European Holocaust Research Infrastructure project, which aims to integrate Holocaust research material from archives across Europe. The main aim of the project will be to make accessible existing Holocaust research collections but the second priority will be to deploy virtual research environments to make use of these resources. D4Science services were seen to support initial requirements well.

References: