Funding: go!digital 2.0 (ÖAW)
Lead: Zentrum für Informationsmodellierung, Universität Graz
Further partner: Know-Center, ISDS, Institut fuer Romanistik der Uni Graz
In the research project “DiSpecs”, researchers analysed the literary works of the “Spectators” from the 18th century with the help of “Distant Reading”. This method enables literary scholars to analyse large quantities of text. In the process, not only the linguistic style of the time, but also ideas, values and moods come to the fore.
The moral weeklies, also called “Spectators”, were a journalistic genre that spread throughout Europe in the 18th century, starting in England. The journals were written in various languages and became an important indicator of the discourse system of the Enlightenment. The contents were aimed at a broad readership. Some focused on theatre and literature, others dealt with science or women’s education. The aim of the journals was to disseminate ideas and values such as morality, virtue and social coexistence through these writings. In this way, they played a decisive role in the formation of opinion at the time.
In “DiSpecs”, researchers from the fields of data science, literary studies and digital humanities worked together. The teams investigated how and which methods proved useful and efficient for analysing the multilingual database. The techniques used included text sentiment analysis, network analysis or topic modelling. Machine learning was used, among other things, to recognise topics or to automatically keyword large amounts of text. The results were then interpreted and validated by the respective departments.
With the help of the methods, it was possible, for example, to identify anonymous authors or to gain information about the attitude of a journal, i.e. whether it has a positive or negative attitude towards certain topics. In the subsequent analysis, certain cultural patterns or coloured opinions were also recognised.
Expertise & technical know-how intertwine
Because of technical gaps among those involved, comprehensive interdisciplinary cooperation is often challenging. While IT experts without the necessary domain knowledge cannot interpret the results of the analyses correctly, literary scientists often lack the technical know-how for the use of machine learning methods. DiSpecs” overcame this barrier and successfully brought together expertise from both fields.
Central multilingual database reads “between the lines“
The machine learning methods are currently being used on a central database for all European weeklies. The aim is to capture and analyse the entirety of the texts. The analyses should make numerous perspectives and narrative forms visible and reduce the workload for literary scholars. In addition, subject scholars can use these methods to gain new insights from this cultural heritage of the Enlightenment period.