In the Tour de CLARIN interview series with prominent CLARIN researchers we talked to Maria Ågren, a professor of history working at Uppsala University, Sweden. She leads the Gender and Work project in which she has collaborated with SWE-CLARIN researchers to create the GaW database, a collection of annotated historical language data that reveal men and women's ways of supporting themselves in the early modern history of Sweden. Jakob Lenardič conducted the interview by e-mail correspondence.
1. Could you please briefly describe your background and tell us what your recent research is about?
I received my first degree in history and Swedish. I also have a diploma for teaching these two subjects in upper-secondary school; however, I never worked as a teacher because I enrolled as a graduate student in history instead. Since 2001, I have been a professor of history and my most recent research has been the Gender and Work (GaW) project that I am leading.
2. How did you get involved with SWE-CLARIN and what impact has this collaboration had on your research?
In the Gender and Work project, we are interested in finding snippets of information about people’s jobs in historical documents, such as farm accounts, diaries, and court protocols. These snippets usually take the form: mend boat, sell eggs, take care of old people, and so on. At an early stage of the project, we told a linguist about our interest in building a database in which this type of information could be stored. She then exclaimed: “Aha! You are interested in verbs!” This comment had two far-reaching effects for our research project. First, we realized that we should call our method verb-oriented because it is a short and efficient way of explaining our approach that everyone immediately understands. Second, this linguist encouraged us to contact Professor Joakim Nivre from SWE-CLARIN which has led to a fruitful collaboration.
3. Which tools and corpora have you used and how did you integrate them into your existing research?
I did not use any existing corpora. Instead, the project has built its own corpus, the GaW database. Project members have gathered and classified thousands of fragments of information from a variety of handwritten historical sources that describe the ways people sustained and provided for themselves. The first stage of the project, which ran between 2010 and 2014 focused on the historical period from 1550 to 1800. The project now continues (from 2017 to 2021) with a focus on the period between 1720 and 1880. The GaW database is accessible to researchers, students, and the general public.
4. Have corpus data helped you reveal any interesting societal and linguistic trends of the periods you are interested in that would have been more difficult to uncover were it not for corpus-based methodology?
Yes, if one accepts my claim that the GaW database is a form of corpus then its data have been absolutely vital to the project. I would even say that most of our results could not have been achieved without it. Likewise, if one accepts that the verb-oriented method is a corpus-based methodology, then the answer is most definitely yes. We have made many interesting discoveries about early modern society.
5. Could you describe the project in more detail? How did the language differ from contemporary Swedish? Are there any interesting differences from a socio-historical point of view? In what way have gendered roles/expectations changed from that time until now?
Gender and Work is a combined research and digitization project at the Department of History at Uppsala University. The aim of the project is to acquire knowledge about the work of both men and women in the past. With the project we have been able to show the importance of the two-supporter model in early modern society; that is, that there was an expectation and practical reality of both men and women contributing to the household’s survival. The project has also shown that what people did for a living in the past had more to do with marital status than with gender. The difference between what married and unmarried people did for a living respectively was larger than the difference between what men and women did for a living respectively.
One could say that early modern gender roles were more similar to the ones we have today – that is, both spouses worked, both spouses were expected to take care of children even if the mother was thought to have a somewhat larger responsibility in this respect, people worked long days and could have to travel far to earn a living – than the ones that developed within the nineteenth-century bourgeoisie.
The Swedish language at the time was of course quite different from modern Swedish. There were no spelling rules, for instance, which makes for varied and, one might say, unorthodox spelling practices. It happened that German words were used in Swedish sentences. For researchers who use the corpus, the language itself is not the largest problem since all scholars involved in the project are historians specializing in the early modern period and they are therefore all accustomed to reading early modern Swedish. The handwriting, on the other hand, is more of a problem; sometimes, the handwriting is so bad that you simply cannot make out what the text is about.
6. Has your field in general embraced the available digital text collections and language technologies? Do Swedish historians make use of language technology or collaborate with research infrastructures such as SWE-CLARIN?
I think the answer to this question must be no. In my opinion the Gender and Work project has been a pioneer within the historical disciplines in Sweden, especially because the collaboration with researchers working with language technology has allowed us to overcome a variety of technical difficulties we faced when dealing with the historical documents.
7. Could you elaborate on these methodological and technical challenges that a researcher working with historical text collections faces with respect to the available infrastructure?
There are two large problems: (1) the inconsistent spelling and (2) the fact that a majority of documents are only available in the original, handwritten form. The former problem is less daunting. In fact, there has been a highly successful collaboration with Professor Joakim Nivre and his now former PhD student Eva Pettersson from SWE-CLARIN, which has led to substantial progress in overcoming the inconsistencies in spelling. For more in-depth information regarding this, see Eva Pettersson’s doctoral dissertation Spelling Normalisation and Linguistic Analysis of Historical Text for Information Extraction.
The latter problem is much more difficult to overcome. In the early modern period, state bureaucracies swelled and this led to a big increase in the production of documents, all of which are valuable to historians. But most of these sources are only available in handwritten form; rarely have they been printed (in which case they can be OCR-read) or digitized directly. If they are not available in digitized form, they cannot be processed automatically. If there existed an easy way of transforming these handwritten documents to digital texts, then the corpus of early modern text material would grow enormously.
Since this is not the case, collaborative interdisciplinary projects like the one between Nivre and Pettersson on the one hand and GaW on the other hand are so rare. In our case, the historians read and annotated the texts manually but at the same time, they also digitized them. This provided Pettersson with the language material on which she could train her normalisation tool. This tool identifies verbs, and particularly verbs describing work activity. The tool is not yet developed to perfection, but hopefully it will one day be possible to run it on digitized texts from the early modern period and in this way speed up the processing of historical texts.
If you are interested in our approach to the extraction of information from historical texts, see “HistSearch – Implementation and evaluation of a web-based tool for automatic information extraction from historical text” by Eva Pettersson, Jonas Lindström, Benny Jacobsson and Rosemarie Fiebranz.
8. What’s your vision for CLARIN 10 years from now? What in your opinion should CLARIN focus on providing?
That it will be a permanent collection of resources and will contain more text corpora from the period between the Middle Ages and ca. 1800. This is the period during which many more documents were produced than in the Middle Ages and most of them were not printed. After around 1800, the handwriting became more similar to the one we have today and more documents were written on typewriters, making them easier to process automatically. The period from 1500 to 1800, on the other hand, is the period that is still largely uncatered for in terms of corpora and text processing tools.
Click here to read more about Tour de CLARIN