Detecting Multiple Transitions in Literary Texts

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 3375–3381 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 3375 Detecting Multiple Transitions in Literary Texts Nuette Heyns, Menno van Zaanen North-West University, Southern African Centre for Digital Language Resources Potchefstroom 2520, South Africa nuette.heyns@gmail.com, menno.vanzaanen@nwu.ac.za Abstract Identifying the high level structure of texts provides important information when performing distant reading analysis. The structure of texts is not necessarily linear, as transitions, such as changes in the scenery or flashbacks, can be present. As a first step in identifying this structure, we aim to identify transitions in texts. Previous work (Heyns and van Zaanen, 2021) proposed a system that can successfully identify one transition in literary texts. The text is split in snippets and LDA is applied, resulting in a sequence of topics. A transition is introduced at the point that separates the topics (before and after the point) best. In this article, we extend the existing system such that it can detect multiple transitions. Additionally, we introduce a new system that inherently handles multiple transitions in texts. The new system also relies on LDA information, but is more robust than the previous system. We apply these systems to texts with known transitions (as they are constructed by concatenating text snippets stemming from different source texts) and evaluation both systems on texts with one transition and texts with two transitions. As both systems rely on LDA to identify transitions between snippets, we also show the impact of varying the number of LDA topics on the results as well. The new system consistently outperforms the previous system, not only on texts with multiple transitions, but also on single boundary texts. Keywords: topic modelling, LDA, transition identification 1. Introduction The digitization and annotation of texts is one of the main focus area in the field of digital humanities. The large number of digitization projects boost the amount of digitally available texts and this, in turn, allows hu- manities scholars to access and analyze texts that pre- viously were difficult to access or process. With the huge amounts of texts available, one may try to identify patterns that cross (large amounts of) texts. However, given the size of the number of texts, close reading approaches (consisting of in depth liter- ary analyses of texts) are practically infeasible. Instead, distant reading approaches (Moretti, 2013), which rely on the automatic analysis of a text, can be considered instead. Distant reading can identify global properties of one or more texts, in contrast to close reading, which focuses on more fine-grained properties (Franzini et al., 2015). One advantage of distant reading is that the computer can perform large scale objective analyses of texts, as opposed to the time consuming, subjective analyses of close reading.1 One task that is particularly useful in a distant reading setting is that of identifying the high level structure of a text. This requires an overview of the entire text (which might be very long, …

Content truncated to 500 words