A Digital History Light approach to using Jupyter Notebook

Much effort is put into integrating digital tools into the regular curriculum of humanities students. However, when introducing digital methods such as ’text mining’ to traditional historians what is not taken into account by instructors is that most historians work with heterogeneous sources and not with a single corpus that can be processed with one 'tool'. A Digital History Light approach may be the way to combine the best of both worlds and sensitise a more heterogeneous group of historians to the merits of computational methods.

I intend to illustrate this with a case study in which a very basic set of Python rules presented in a Jupyter notebook was used to teach a small group of classical not digitally geared historians (a master student, a PhD and a lecturer) to apply text mining to a database of 700 veteran memoirs. My experience with this experiment and the many workshops that I attended and led, is that if you want to teach students to combine close and distant reading and quantitative and qualitative analysis, it is fruitful to breakdown the standard repertoire of a tool, and cherry pick single elements that seamlessly connect to the standard workflow of a traditional historian. As a preparation, it can have an added value to provide literature that reflects on the relation between language and historical knowledge in a qualitative manner. It can tweak the mind of the historian to a linguistic frame, before the computational dimension is introduced. At the same time, the computational geared scholar should understand that for the historian particular terms function as cues to what people at the time actually did. How were women treated? What kind of violence was permitted? How did they spend their free time? These cues point to contexts that may have left other traces, which then can be triangulated with the passages found in the database. One can think of military reports, letters and photos. It is this heterogeneous use of sources, not the focus on patterns within a particular corpus, which characterises most of the scholarly work of historians and therefore tends to be at odds with a computational approach. Through this 'Textmining Light' approach, I want to show how to combine the best of both worlds in a time frame that is manageable.