Tour de CLARIN highlights prominent User Involvement (UI) activities of a particular CLARIN national consortium. This time the focus is on Denmark and Klaus Nielsen, the chief editor at the Grundtvig Study Centre. The interview was conducted via Skype by Jakob Lenardič.
1. What is your scholarly background and your current academic position?
I obtained my PhD from the University of Copenhagen in 2012 and my thesis was a combination of traditional literary theory and book history, a philological field that focuses on a more mechanical-analytical study of the publication process of literary works. I focused on Gittes monologer, a famous collection of satirical poems by the Danish poet Per Højholt published in different versions between 1980 and 1984. I was able to observe crucial textual differences between their various published versions, which allowed me to arrive at a much richer interpretation of the poems that wouldn’t be possible with the final, best-known 1984 version alone. This showed me how important it is to combine traditional qualitative literary analysis with analytical methods that also take into consideration non-textual information such as publication history.
I now work as chief editor at Grundtvig Study Centre, where we are preparing a critical edition of the collected works of N.F.S. Grundtvig, a very prolific and multidisciplinary Danish author who published around 37,000 pages of text from 1804 to his death in 1872. We are making this corpus available in an online environment, with manual annotations that follow the scholarly standards of textual criticism. In a sense, my PhD was an important methodological steppingstone for my current work related to the Grundtvig’s Works Corpus, which also involves a close study of the differences between the various published editions.
2. The Grundtvig’s work corpus has been published through the CLARIN-DK repository. How did this collaboration start? How do you benefit from this collaboration?
We released the first version of our corpus through the CLARIN-DK repository in 2018 at the suggestion of Lene Offersgaard, with whom we were collaborating on a related project at the time. This was a great opportunity for us because we had been receiving feedback from some of our more devoted users who said they wanted the corpus in a downloadable format. We’ve also made an agreement with CLARIN-DK that as soon as we publish a new version of the corpus through our online environment, we’ll also update the version deposited in the repository with the newest, more richly annotated one.
3. How is Grundtvig’s corpus structured? What are some of the challenges you come across when annotating the corpus?
The corpus is extremely varied in terms of content, since Grundtvig was a polihistorian who wrote on a variety of different subjects. Perhaps most prominently, he wrote books on Danish history and Nordic mythology, carried out linguistic studies of Old Icelandic and Old English, translated from Latin, wrote political and philosophical texts, and composed around 1,500 hymns, many of which are still sung today in Denmark. For this reason, Grundtvig’s views are representative of the intellectual and cultural zeitgeist of Denmark in the 19th century.
There’s a downside to his varied repertoire, in that annotation is still manually intensive. We do use a database for place and person names that we feed into a named-entity recognizer, but even in this case, we often have to manually verify the results. For example, Grundtvig often refers to the philosopher Søren Kierkegaard, who was a contemporary of his, and our software is generally successful in identifying this particular named entity. However, Grundtvig often refers to him by his last name only, but since Søren Kierkegaard had a brother who was also a published author in the same period, we have to manually check the automatic recognition to make sure that the software made a link to the correct referent. In addition to this, we often come across obsolete words, in which case we manually add their possible historical meaning. This can only be done by closely reading and interpreting the surrounding text. Nevertheless, we will use the parts that have already been annotated as a baseline for a semi-automated processing of the remaining two-thirds of the corpus in the future.
One of the greatest challenges in terms of mark-up pertains to identifying Biblical references, especially in cases where Grundtvig doesn’t use direct quotes taken from the Bible but his own modified variants, or where he makes indirect references to the more obscure motifs and quotes. Although we have theologians both internal and external who closely read the texts and manually identify such references, it would be invaluable if we could also make use of a language tool that would help automatize this process of identification. I don’t think that such a tool exists yet, but it would be a very welcome addition to the CLARIN infrastructure in my opinion. Similarly, it would be great to have a tool that can automatically recognize proverbs and sayings, which abound in Grundtvig’s works, given that his work is a major part of the Danish cultural heritage. Although I’m not an expert in digital technologies, it seems that developing such a tool wouldn’t be too hard a task, as there already exist readymade digital collections of Danish proverbs that could be used as a baseline for training the tool.
4. Has the corpus been successfully used by an external research project?
Yes, Baunvig and Nielbo (2017) have used our corpus in a case study to determine how digital methods can benefit the analysis of very large collections of written text, and uncover new perspectives and interpretations. Grundtvig Studies is a popular subfield in literary history in Denmark, and many studies on Grundtvig have been published in the past fifty years. However, previous researchers weren’t able to use digital methods and tools, which means that their claims were influenced by the limitations inherent to a purely manual approach to analysis. As I’ve said, Grundtvig produced around 37,000 pages in his lifetime, which is simply too much text for an individual researcher to read and then be able to recollect the finer details. For instance, there is an older study in which it is claimed that Grundtvig started suffering from a series of psychological problems in the 1830s, which was reflected in the texts he wrote in this decade. However, Baunvig and Nielbo (2017) were able to show, by using quantitative methods such as measuring the amount of information entropy in the corpus, that his psychological turmoil actually started earlier than was previously claimed, which is of course an important finding from a purely historical viewpoint. There has also been a follow-up study of our corpus conducted by Nielbo et al. (2018).
5. What makes this corpus particularly valuable for the CLARIN infrastructure?
I think that our rather thorough manual approach to the corpus is an important contribution for a more accurate understanding of the historical developments of the Danish language, especially its orthography. What is important in this respect is that there were no orthographic rules in Grundtvig’s time, only tendencies, which means that spelling was quite liberal in comparison to contemporary Danish. Consequently, we’re often in doubt whether the way Grundtvig spelled a certain word is an instance of spelling variation that was attested at the time or if it is just a spelling mistake on his part. This is particularly problematic in cases where Grundtvig’s idiosyncratic spelling can’t be found in the historical dictionaries of 19th century Danish, since this intuitively makes you think that the spelling variant was a mistake. However, such dictionaries weren’t compiled on the basis of the original edition but often used later published editions that had gone through the editing process, where spelling variation was normalized. This means that if a researcher wanted to study the vocabulary of 19th century Danish just on the basis of such dictionaries, he or she would miss the attested variations and consequently get a warped view of how people actually wrote at the time. By contrast, we spend a lot of time closely analysing and proofreading the materials, so we are able to present a resource that serves as a much more complex, as well as accurate, presentation of the linguistic situation at the time.
6. Could you give an example of such orthographic variation? How did you resolve it?
I actually came across a fairly interesting orthographic problem just recently when I was annotating Grundtvig’s History of the Northmen, which is one of the few texts he had written in English. In this text, Grundtvig used the word kempion in the sense of “champion” or “hero”; however, this spelling variant isn’t listed in the Oxford English Dictionary, which only includes the variant campion with an a instead of an e. Because my colleagues and I weren’t sure how to solve this issue, we consulted a Professor of Middle English, and he believed it to be a spelling mistake that should be corrected in the edited corpus, given that the Oxford English Dictionary is extremely comprehensive and thorough in its account of English etymology. However, when I searched for the variant kempion on Google, I found out that it was actually attested at the time, and it was for instance used by Sir Walter Scott in his 1822 novel The Pirate, which Grundtvig was alluding to.
7. Are there any other aspects of the CLARIN-DK infrastructure that are important for your work at the centre?
Yes, especially in relation to how proactively they reach out as part of their user-involvement initiative. Last year, CLARIN-DK organized a tutorial for the philologists at our centre where they demonstrated how Voyant tools can simplify our annotation process. Using Voyant has turned out to be extremely helpful when we come across obsolete phrases the meaning of which we don’t know and can’t find in the historical dictionaries. By using Voyant’s extended search capabilities and visualisation tools, we are now able to easily chart the occurrences of this unknown phrase in the entire corpus, and then extract only those texts where this phrase seems to occur in a similar context, which then helps us determine its actual meaning.
I am also pleased to say that CLARIN-DK has already made the first version of our corpus available through their installation of the Voyant tools. We plan on updating this test version with newer ones with regularity. In the long run, I believe the availability of the corpus through CLARIN-DK’s Voyant tools will significantly streamline user assistance.
8. Your professional website says that you’re also interested in audio literature. Is this something that you’re still actively researching?
No, my research on audio literature was mostly confined to my PhD project, because Per Højholt, who is the author of the poems that I was analysing, had read them aloud on Danish radio in the 1980s. By using an audio-analysis software called PRAAT, I measured prosodic features such as the author’s pitch and reading speed, and I was able to see how he deliberately changed his voice in accordance with the way the point-of-view character developed through the course of the poems’ narrative. This was a rather small but important finding since it hadn’t been previously acknowledged in the relevant literature on Gittes Monologer how the author’s spoken performance of his own work added new dimensions to the understanding of the poems themselves.
9. What kind of new research questions does audio literature offer in the context of Digital Humanities? Do you think that CLARIN could contribute to this field?
When I was writing my thesis, research on audio literature was still a very new field, but nowadays it is more readily agreed upon that audio recordings can serve as crucial material for textual analysis. Literary theorists are now conducting important research on the link between the reader of the audio text and the content of the text itself, and this opens up many interesting questions. Let’s say, for instance, that we are dealing with a novel written in the first person, and that the narrator is a woman. Should the reader of the audio version then also be a woman, or conversely, what interpretative repercussions would arise if the reader were actually a man? That is, the person’s voice crucially affects the way people perceive the text, much in the same way that the sort of typography of an old book can evoke various pre-conceptions in the reader about the book’s content.
Given how audio literature opens up interesting questions relevant for the emerging digital humanities, I think that new digital tools for analysing recorded literary works would serve as very welcomes additions to the CLARIN infrastructure.
10. What are your hopes for CLARIN-DK in the future?
I think that one of the future challenges for Digital Humanities in Denmark is to find a common platform where our whole research community can have a more unified and interoperable access to as many carefully annotated resources as possible. I believe that CLARIN-DK is an excellent candidate in the country for this, because our experience with releasing the Grundtvig’s Work corpus has proven to us that their repository is a stable environment through which corpora can be released in a sustainable fashion and with well-presented metadata. On top of that, the repository also allows us to integrate our corpora with other services in the consortium. For this reason, it can only be a good thing if more digital humanities scholars in Denmark decide to deposit their resources in the CLARIN-DK repository.
Click here to read more about Tour de CLARIN