The Researcher-in-residence programme at the KB, National Library of the Netherlands

The KB, National Library of the Netherlands, started its mass digitisation projects in 1999, having digitised selected items from its collection since 1995. With currently more than 16 million pages of newspapers, books, journals and radio bulletins available online at the search portal Delpher¹ the digital collection of the KB grew to a size where researchers could not close read² all documents anymore.

Digital methods

More and more research was being done using digital tools and methods, such as ngram viewers³ and linked search interfaces⁴. However, this work was mostly organised outside of the KB by researchers. Projects where the KB was involved in as a partner, such as IMPACT⁵, and CATCH⁶, were dwindling down due to cuts in governmental and European funding.

New ways of collaboration

We felt that we were losing the oversight of how researchers worked with our digitised collection and wanted to find a way to strengthen our relationship with the research community and to learn about the new methods in humanities research and how these worked in relation to our own collection. To do this, we set up our Researcher-in-residence programme in 2014.

Setup of the programme

In this programme, we invite early career researchers to propose a project, following a call for proposals⁷, that requires computational techniques to answer a research question using our digital collection. The projects can be six months maximum and the researchers are seconded to the KB for 50 percent of their working time. During the projects, they are supported by the KB Digital Humanities team, consisting of a curator Digital Collections, two Digital Scholarship Advisors and two research software engineers, of which one supports the researcher for one day a week. We can house two researchers a year, ideally without much overlap, and ask them to spend at least one day a week in the KB at the Research department.

Requirements

The projects and researchers need to fulfil several requirements in order to be accepted into the review process. First, the researcher needs to be at the beginning of their career. We define this by being in the final phase of a PhD project, or having received their PhD in the last five years. Also, due to practical reasons, the researcher needs to be employed by a European research institute as a teacher or researcher, be available for six consecutive months and have the possibility of travelling to the KB one day a week.

Application process

The projects are proposed using a form⁸, unifying the proposals as much as possible. All entries are judged on:

• Originality and quality
• Link with techniques and methods currently applied at the KB Research Department
• Feasibility (technically, legally and practically)
• How the KB data will be showcased and used
• Wether the end results are of use for a wider community

We recommend applicants to contact us before submitting to discuss their proposal, in order to line up their project with the goals and criteria of the programme and the KB.

Review process

The projects are evaluated by a review committee of 10-14 reviewers, consisting of experts in various fields, such as history, computer science and literary studies, with one expert as the chair. The KB also contributes to this committee with two votes, one from the Research department, where the work is being carried out, and one from the Collections department, where the collections are housed on which the work is being built. Two projects are selected during this meeting, the availability of the researchers taken into account, and two back up projects in case one researcher becomes unavailable.

Example projects and outcomes

The projects we ran have ranged from researching possibilities of working with topic modelling software, extracting more than 1.5 million images from the digitised newspapers⁹, to automatically recognising genre in newspaper articles¹⁰. All projects are eventually shared through the KB Lab¹¹ with live demos where possible. In one of our latest projects we worked together with Melvin Wevers¹², a PhD student from Utrecht University. He is interested in the depiction of objects in newspaper advertisements and wanted to know how he could automatically recognise items such as cigarettes, cars and toothpaste. Together with Juliette Lonij¹³ they made SIAMESE¹⁴ - a tool to browse almost 500.000 newspaper advertisements¹⁵using image recognition software. The tool is freely available through our Lab and on Github¹⁶, and the dataset is available for research purposes upon request.

Lessons learned: Setup

We have run the programme for almost four years now, with nine researchers so far and have also made some necessary changes to the setup. To start, we initially ran three projects in a pilot phase in 2014-2015, where we worked with researchers we had invited to the KB. However, we felt that using a call for proposals would reach a wider group of researchers, allowing a more transparent process and it also gave us the opportunity to connect with the experts we invite into the review committee. During these pilots, we also learned that three projects in one year were too much. They had some overlap in time, which was interesting for the researchers to network and discuss ideas, but proved to be too impractical for the KB team. This led to the decision to invite just two researchers per year and keep the overlap to a minimum.

Lessons learned: Review process

The review process has also been adjusted during the years and the form underwent several changes, ending up as a Word document with several sections where each section has a maximum amount of words to be used and includes a short CV of the proposer, their top three publications and a bibliography. The reviewers are asked to fill out the comments in a dedicated form which rates the criteria listed above on a 1 to 5 scale. Consequently, we ask the reviewers to make a ranged list of the projects to discuss during a review meeting at the KB. This facilitates the discussion and gives a quick overview of which projects are most promising and where more discussion is needed.  

Future

The programme was originally set up to run for four years. This means that the programme is to be evaluated in 2018 and a decision will be made to continue it or not. We will also run two projects in 2018, both of which have been selected in the review meeting in September 2017. We feel that the programme has brought us what we hoped it would and would encourage other libraries with similar wishes and collections to undertake such an endeavour, as it not only strengthens your relationship with the research community, it also brings you valuable knowledge about tools, methods and your own collection.