Fabula-NET is an interdisciplinary collaboration of researchers combining knowledge from literary studies, computational linguistics, machine learning, and cognitive science seeking to clarify our understanding of literary quality. The aim of the research project is to develop computational tools for quality assessment and automated narrative analysis of literature for the benefit of both editors, writers, librarians, and researchers.
A key objective for the research group is to develop text classification models widely available for specialists and professionals: In the publishing industry, models developed to evaluate large amounts of texts and point out the high-quality ones will make screening processes significantly more efficient as editors will be able to focus on the manuscripts with most potential according to the models. Similarly, writers can use the models to evaluate their own work, whereas librarians can apply the models to optimize search processes and text assessment. For researchers within literary studies, the models will enable comparisons of large corpora of world literary texts on a much larger scale.
The classification tools are developed based on three distinct parameters for success: firstly, the intrinsic success, which denotes the inner structure of a literary work and is important in terms of quality and success as certain narrative structures tend to characterize successful narratives. Secondly, the research group applies an extrinsic parameter of success evaluating a given literary work in terms of literary reviews and popularity at libraries. The last and third parameter is termed the preferential succes as it evaluates the succes of the literary work among different age groups and social groups in order to point out any potential preferences among certain segments.
The research group combines fractal analysis, sentiment analysis and advanced language models with deep neural networks and machine learning. Thus, based on the parameters above, they are able to train a system able to evaluate the quality of new literary works based on the data is has been trained on.
The scope of the Fabula-NET project depends entirely on interdisciplinary collaboration between the computational field and the humanities, AI and literary studies. As such, the project embodies one of the central goals of Center for Humanities Computing (CHC), which is to support the use and development of computational approaches and digital data within the humanities.
The Fabula-NET research group consists of five members representing several disciplinary backgrounds, all connected to CHC: PI Kristoffer Laigaard Nielbo is Associate Professor, head of CHC and specialized in data-driven approaches to sociocultural systems, whereas Co-PI, Mads Rosendahl Thomsen, is Professor of Comparative Literature and has a special interest in canonization of literature and the great unread. The group furthermore consists of Yuri Bizzoni, computational literature analyst and post.doc. at CHC, Ida Marie S. Lassen, Ph.D. candidate at CHC and interested in critical approaches to data and model development (particularly in the case of gender biases), and Telma Peura, research assistant and Master Student in Digital Humanities at Helsinki University.
One of the noteworthy outcomes of the Fabula-NET project so far is a recent study revealing that Danish literary criticism is characterized by a statistically significant gender imbalance. The study reveals a recurrent and structural imbalance as both male and female reviewers tend to assign the highest grades to male authors in newspaper criticism. Moreover, the study shows that 2/3 of the Danish newspaper reviewers are male and that newspaper reviews of male authors are far more frequent than newspaper reviews of female authors. Learn more and read the report based on the study and the other outcomes of Fabula-NET research here.
Fabula-NET is funded by a grant of DKK 5.3 million from the Velux Foundation
Collaborate with our Research Software Engineers, Data Scientists or Data Managers