The National Library of Finland (NLF) has done long-term work to digitise and make available our unique collections. The digitisation policy defines what is to be digitised, and it aims not only to target both rare and unique materials but also to create a large corpus of certain material types. However, as digitisation resources are scarce, the digitisation is planned annually, where prioritisation is done. This involves the library juggling the individual researcher needs with its own legal preservation and availability goals. The digital presentation system at digi.nationallibrary.fi plays a key role, since it enables fast operation by being next to the digitisation process, and it enables a streamlined flow of material via a digital chain from production and to the end users.

In this paper, we will describe our digitisation process and its cost-effective improvements, which have been recently applied at the NLF. In addition, we evaluate how we could improve and enrich our digital presentation system and its existing material by utilising results and experience from existing research efforts. We will also briefly examine the positive examples of other national libraries and identify universal features and local differences.

Ømålsordbogen (the Dictionary of Danish Insular Dialects, henceforth DID) is an historical dictionary giving thorough descriptions of the dialects, i.e. the spoken vernacular of peasants and fishermen, on the Danish isles Seeland, Funen and surrounding islands. It covers the period from 1750 to 1950, the core period being 1850 to 1920. Publishing began in 1992 and the latest volume (11, kurv-lindorm) appeared in 2013 but the project was initiated in 1909 and data collection dates back to the 1920s and 1930s. The project is currently undergoing an extensive process of digitization: old, outdated editing tools have been replaced with modern (database, xml, Unicode), and the old, printed volumes have been extracted to xml as well and are now searchable as a single xml file. Furthermore, the underlying physical data collections are being digitized.

In the following we give a brief account of the digitization process, and we discuss a number of questions and dilemmas that this process gives rise to. The collections underlying the DID project comprise a variety of subcollections characterized by a large heterogeneity in terms of form as well as content. The information on the paper slips are usually densified, often idiosyncratic, and normally complicated to decode, even for other specialists. The digitization process naturally points towards web publication of the collections, either alone or in combination with the edited data, but it also gives rise to a number of questions. The current digitization process being very basic, only adding very few metadata (1-2 or 3), we point to the obvious fact that web publication of the collections presupposes an addition of further, carefully selected metadata, taking different user needs and qualifications into account. We also discuss the relationship between edited and non-edited data in a publication perspective. Some of the paper slips are very difficult to decipher due to handwriting or idiosyncratic densification and we point out that web publication in a raw, i.e. non-edited or non-annotated form, might be more misleading than helpful for a number of users.

This presentation will focus on the re-use of data relating to collections in libraries, museums and archives to address research questions in the humanities. Cultural heritage materials held in institutional collections are crucial sources of evidence for many disciplines, ranging from history and literature to anthropology and art. They are also the subjects of research in their own right – encompassing their form, their history, and their content, as well as their places in broader assemblages like collections and ownership networks. They can be studied for their unique and individual qualities, as Neil McGregor demonstrated in his History of the World in 100 Objects, but also as components within a much larger quantitative framework.

Large-scale research into the history and characteristics of cultural heritage materials is heavily dependent on the availability of collections data in appropriate formats and sufficient quantities. Unfortunately, this kind of research has been seriously limited, for the most part, by lack of access to suitable curatorial data. In some instances this is simply because collection databases have not been made fully available on the Web – particularly the case with art galleries and some museums. Even where databases are available, however, they often cannot be downloaded in their entirety or through bulk selections of relevant content. Data downloads are frequently limited to small selections of specific records.

Collections data are often available only in formats which are difficult to re-use for research purposes. In the case of libraries, the only export formats tend to be proprietary bibliographic schemas such as EndNote or RefCite. Even where APIs are made available, they may be difficult to use or limited in their functionality. CSV or XML downloads are relatively rare. Data licensing regimes may also discourage re-use, either by explicit limitations or by lack of clarity about terms and conditions.

Even where researchers are able to download usable data, it is very rare for them to be able to feed back any cleaning or enhancing they may have done. The cultural heritage institutions supplying the data may be unable or unwilling to accept corrections or improvements to their records. They may also be suspicious of researchers developing new digital services which appear to compete with the original database.

As a result, there has been a significant disconnect between curatorial databases and researchers, who have struggled to make effective use of what is potentially a very rich source of computationally usable evidence. One important consequence is that re-use of curatorial data by researchers often focuses on the data which are the easiest to obtain. The results are neither particularly representative nor exhaustive, and may weaken the validity of the conclusions drawn from the research.

Some recent “collections as data” initiatives (such as collectionsasdata.github.io) have started to explore approaches to best practice for “computationally amenable collections”, with the aim of “encouraging cultural heritage organizations to develop collections and systems that are more amenable to emerging computational methods and tools”. In this presentation, I will suggest some elements of best practice for curatorial institutions in this area.

My observations will be based on three projects which are addressing these issues. The first project is “Collecting the West”, in which Western Australian researchers are working with the British Museum to deploy and evaluate the ResearchSpace software, which is designed to integrate heterogeneous collection data into a cultural heritage knowledge graph. The second project is HuNI – the Humanities Networked Infrastructure – which has been building a “virtual laboratory” for the humanities by reshaping collections data into semantic information networks. The third project – “Reconstructing the Phillipps Collection”, funded by the European Union under its Marie Curie Fellowships scheme – involved combining collections data from a range of digital and physical sources to reconstruct the histories of manuscripts in the largest private collection ever assembled.

Curatorial institutions should recognize that there is a growing group of researchers who do not simply want to search or browse a collections database. There is an increasing demand for access to collections data for downloading and re-use, in suitable formats and on non-restrictive licensing terms. In return, researchers will be able to offer enhanced and improved ways of analyzing and visualizing data, as well as correcting and amplifying collection database records on the basis of research results. There are significant potential benefits for both sides of this partnership.