Tantrums and Traitors: a Diachronic Analysis of Emotions in Parliamentary Debates on War Criminals and Collaborators

In this study discussions about the punishment of collaborators and war criminals in the Netherlands are investigated by analysing the verbatim minutes of parliament. The application of text mining techniques to this digitised historical text corpus allows for a diachronic perspective on a historical case study. With this paper, I present the historical case, the materials and techniques used, and some insights based on preliminary results. I will also address general advantages and limitations of using text mining in historical research. Aim of this investigation is to explicitly investigate and discuss the validity of the use of emotion lexicons in diachronic historical research.

Historical case
After the Netherlands was liberated in May 1945, the issue of the punishment of collaborators and war criminals became acute. At peak, during the summer of 1945, more than hundred thousand suspected people were kept in custody within the Netherlands. For contemporary politicians, this issue was pressing and demanded quick action. Heated debates in Dutch parliament followed. This was, however, not the end of the story. For the next five decades, the question caused recurring heated political controversies. The debates in Dutch parliament about the punishment, penalty reduction, or release of these people are not only among the longest debates in Dutch parliamentary history, but are generally considered to have been the most emotionally charged (Bootsma and Griensven, 2003; Tames, 2013).
Controversies kept recurring in 1952, 1966, 1972, and 1989. The 1952 controversy started when Queen Juliana proposed (by royal decree) to turn the death sentence of German war criminal Willy Lages into life imprisonment. This proposal caused lengthy debates in parliament and mass demonstrations in Amsterdam, attended by approximately 15.000 people. This manifestation of public anger stands in apparent contrast with the common assumption in Dutch historiography about ‘the silence of the 1950s’. This silence refers not only to talking about past war experiences, but also to (not) expressing related emotions (Withuis, 2002).
Decades later the issue proved to still have the potential to cause controversies. By 1972 only three foreign war criminals were still incarcerated in the Dutch city of Breda. Public unrest rose when the potential release of these last three German war criminals in Dutch custody was discussed in parliament. Newsreels show the manifestations of sadness and anger of the – again – more than 15.000 people that demonstrated in- and outside the parliamentary building in The Hague. This Breda Three controversy became one of the best-known emotional Second World War-related affairs in the Netherlands (Piersma, 2005). The Minister of Justice at the time, Dries van Agt, called this debate the most emotional episode of his entire political career (Bootsma and Griensven, 2003).

Questions
This presumed high emotional charge of well-known controversies seems to be taken for granted. This leaves historians with not only substantive but also epistemological questions. To start with the latter: how can we observe, interpret, and compare emotionality in the context of historical research? Investigating emotionality is full of complexities. First, emotions are volatile and thus difficult to grasp – especially in retrospect. Next, it is to be expected that historians reading parliamentary proceedings, trying to identify elusive emotionality, are influenced by their own preconceptions. This is especially the case in ethically charged controversies in the relative recent past. By approaching a well-researched case historically with an innovative text mining approach, I hope to be able to also investigate these personal judgments of emotion in historiography.
This study aims to answer the following substantive historical questions: how emotional were these debates? Which emotions were distinctive? How does the 1952 parliamentary discussion and the uproar it caused relate to the generally assumed ‘silence of the 1950s’? How do the different (war-related) discussions relate to each other? How exceptional are the war criminal debates when they are compared to other contemporary issues?

Methodology
This investigation uses emotional word use to investigate emotions present in the discussions. I rely on generic emotion lexicons that are produced outside the field of history. They are derived from the fields of computational linguistics and psychology (Boot et al., 2017; Mohammad and Turney, 2013; Pennebaker, 2013; Tausczik and Pennebaker, 2010). The emotion lexicons used are generic in the sense that that their creation is not based on a single (type of) dataset. Using lexicons to identify, analyse, and evaluate manifestations of emotions, can be considered as a rather crude method. On the level of a single particular sentence, these lexicons may not always be reliable in determining whether a certain emotion is manifest. However, when these lexicons are applied to large corpora, they are reliable in determining whether a certain part of the corpus has more (or less) manifestations of a certain emotion when compared to others (Mohammad, 2012). Taking into account the ambition to address the influence of personal judgment in this investigation, I consider it an important advantage that the lexicons are from outside the specific domain of this study, and not influenced by personal judgments of the investigator(s) towards the historical sources or themes under scrutiny. The use of emotion mining as a method in historical research forces researchers to formalize what they want to know (what is emotional), and formalize how their analytical process leads to their answers and conclusions (Rockwell and Sinclair, 2016).
The historical parliamentary debates are analysed by using open-source text mining packages (e.g. quanteda and tm) within the R-programming environment (Benoit et al., 2018; Feinerer et al., 2008; Feinerer and Hornik, 2018). The digitised parliamentary debates are pre-processed: they are lemmatised and stopwords and interpunctions are removed (Marx et al., 2012). Next, the corpora are transformed into Document Term Matrices (DTM). The lexicons are used to score the parliamentary records on manifestation of words relating to different categories of emotion (e.g. anger, sadness, trust, fear). In this process, a Term Frequency – Inverse Document Frequency (TF-IDF) weighting is assigned to each unique word in the corpus. The DTMs are used to measure how distinctive words from an emotion category (from the lexicons) are for a specific document relative to all other documents. The output is then turned into boxplots and distribution graphs that are used for statistical evaluation and comparison of the results.
This output generates insights in which different emotional words were distinctive, and in which proportions they were manifest in the parliamentary discussions. Outputs are compared over time to come up with a long-term perspective on historical developments in discussing collaborators and war criminals in Dutch parliament. These quantitative results are also confronted with more traditional historical analysis of both primary and secondary sources.

Bibliography

Benoit, K., Watanabe, K., Wang, H., Nulty, P., Obeng, A., Müller, S. and Matsuo, A. (2018). quanteda: An R package for the quantitative analysis of textual data.
Journal of Open Source Software,
3(30): 774 doi:10.21105/joss.00774.

Boot, P., Zijlstra, H. and Geenen, R. (2017). The Dutch translation of the Linguistic Inquiry and Word Count (LIWC) 2007 dictionary.
Dutch Journal of Applied Linguistics,
6(1): 65–76 doi:10.1075/dujal.6.1.04boo.

Bootsma, P. and Griensven, P. van (2003). `Teleurstelling is mijn opperste emotie’: Vragen over emotie in de politiek aan A.A.M. van Agt. In al, C. van B. et (ed),
Jaarboek Parlementaire Geschiedenis, 2003. Emotie in de Politiek. Den Haag: SDU Uitgevers.

Feinerer, I. and Hornik, K. (2018).
Tm: Text Mining Package. https://CRAN.R-project.org/package=tm.

Feinerer, I., Hornik, K. and Meyer, D. (2008). Text Mining Infrastructure in R.
Journal of Statistical Software,
25(5): 1–54.

Marx, M., Doornik, J. V., Nusselder, A. and Buitinck, L. (2012). Dutch Parliamentary Proceedings 1814-2012, non-semanticized Data Archiving and Networked Services (DANS) doi:10.17026/dans-xk5-dw3s. https://easy.dans.knaw.nl/ui/datasets/id/easy-dataset:51640 (accessed 28 February 2019).

Mohammad, S. M. (2012). From once upon a time to happily ever after: Tracking emotions in mail and books.
Decision Support Systems,
53(4): 730–41 doi:10.1016/j.dss.2012.05.030.

Mohammad, S. M. and Turney, P. D. (2013). Crowdsourcing a Word-Emotion Association Lexicon.
Computational Intelligence,
29(3): 436–465.

Pennebaker, J. W. (2013).
The Secret Life of Pronouns: What Our Words Say about Us. New York: Bloomsbury Press.

Piersma, H. (2005).
De Drie van Breda: Duitse Oorlogsmisdadigers in Nederlandse Gevangenschap, 1945-1989. 1st ed. Amsterdam: Balans.

Rockwell, G. and Sinclair, S. (2016).
Hermeneutica: Computer-Assisted Interpretation in the Humanities. Cambridge, Massachusetts and London: The MIT Press.

Tames, I. (2013).
Doorn in het vlees: foute Nederlanders in de jaren vijftig en zestig. 1st ed. Amsterdam: Balans.

Tausczik, Y. R. and Pennebaker, J. W. (2010). The Psychological Meaning of Words: LIWC and Computerized Text Analysis Methods.
Journal of Language and Social Psychology,
29(1): 24–54 doi:10.1177/0261927X09351676.

Withuis, J. (2002).
Erkenning: van oorlogstrauma naar klaagcultuur. Amsterdam: De Bezige Bij.