Is Expert Knowledge Key? Scholarly Interpretations as Resource for the Analysis of Literary Texts in Computational Literary Studies

Robert Jäschke & Steffen Martus

Starting with “Key passages in literary works” in the first phase of our project within the SPP 2207 „Computational Literary Studies“ (CLS), we have explored new ways to draw on expert knowledge in literary studies, expressed in interpretative texts. In this way, we contribute to the goal stated in the research program of the SPP: to combine research results acquired with qualitative methods with quantitative methods. In this second phase, we extend this approach, to open up further possibilities for re-using existing resources of literary studies and for combining research practices, which are common in literary studies and in CLS. As we have shown in the analysis of key passages, where we have empirically recorded which passages professional interpreters consider particularly important, scholarly interpretations set weighted priorities or focuses, whereas CLS typically proceed in an equally selective and focusing manner. Standard quantitative clustering approaches assume that literary texts are evenly significant in all their passages (in the respect that was prioritized for the investigation, for example, most frequent words, emotions etc.). Taking the analysis of key passages as a starting point, we now want to ask to what extent expert knowledge can be key to these weighting procedures of interpretation, and how such ‘skillful’ readings can advance the approaches of CLS.While key passages remain partly in the focus of the project, we aim to leverage expert know-how in literary studies to improve CLS methods in three different ways. The foci have emerged from our previous research, so that the studies pragmatically mutually benefit from each other; each of them connects to existing CLS expertise to test riskier research options from there. 1. Narrative structure detection: We want to ask how the determination of events can contribute to the understanding of key passages and how narratological aspects, especially concerning plot structures in literary texts, can be identified. 2. Sentiment analysis: We want to reverse the perspective of common sentiment analysis approaches and leverage the already existing knowledge about emotions in interpretative texts. Last but not least, we want to ask how text-based approaches for emotion detection can be combined with those that presuppose knowledge that clearly goes beyond the emotions thematized. 3. Text clustering: Finally, we turn to complex clustering using the example of grouping literary texts into literary epochs. At the same time, we want to ask how what we have heuristically called evenly (CLS) or weighted selective/focused (experts) text processing can be combined. To sum up, our overall research question is: How can already established interpretative knowledge be used in new ways within the framework of CLS in order to make the most effective use of existing resources in literary studies and to make interdisciplinary connections available?