SLäNDa version 2.0: Improved and Extended Annotation of Narrative and Dialogue in Swedish Literature

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 5324–5333 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 5324 SL¨aNDa Version 2.0: Improved and Extended Annotation of Narrative and Dialogue in Swedish Literature Sara Stymne and Carin ¨Ostman Department of Linguistics and Philology, Department of Nordic Languages Uppsala University, Sweden sara.stymne@lingfil.uu.se,carin.ostman@nordiska.uu.se Abstract In this paper, we describe version 2.0 of the SL¨aNDa corpus. SL¨aNDa, the Swedish Literary corpus of Narrative and Dialogue, now contains excerpts from 19 novels, written between 1809–1940. The main focus of the SL¨aNDa corpus is to distinguish between direct speech and the main narrative. In order to isolate the narrative, we also annotate everything else which does not belong to the narrative, such as thoughts, quotations, and letters. SL¨aNDa version 2.0 has a slightly updated annotation scheme from version 1.0. In addition, we added new texts from eleven authors and performed quality control on the previous version. We are specifically interested in different ways of marking speech segments, such as quotation marks, dashes, or no marking at all. To allow a detailed evaluation of this aspect, we added dedicated test sets to SL¨aNDa for these different types of speech marking. In a pilot experiment, we explore the impact of typographic speech marking by using these test sets, as well as artificially stripping the training data of speech markers. Keywords: Literary dialogue, Narrative, Direct speech, Annotation 1. Introduction Modern computational models for text analysis have al- lowed large-scale studies in many fields of research, such as language studies and literary studies. Early studies in computational literary studies, as well as in other application areas, often used unsupervised methods, such as topic modelling (Boyd-Graber et al., 2017). However, for many tasks, supervised methods are preferred, necessitating the creation of annotated corpora. In this work we present a corpus, SL¨aNDa, which contains annotations of speech, narrative, and other non-narrative parts, as well as mentions of speak- ers, in Swedish literature from the 19th and 20th cen- turies. The period around the turn of the 19th cen- tury is a period of modernization of the Swedish writ- ten language, where literature, and especially literary dialogue, is believed to have played a role (Teleman, 2003). SL¨aNDa is the only available large-scale corpus of Swedish literature annotated for narrative, speech, and speakers. While it may seem trivial to extract speech segments in a language like English, where speech is typically marked with quotation marks, this is not the case for all languages. In Swedish there are two main types of speech marking, either with quotation marks, or with a dash at the start of a speech segment, not marking where the speech segment ends, where the speech tag starts, or where the speech continues after the speech tag. Example (1) shows an example from Strindberg, page 138, which starts with the speech segment ‘My ladies’, marked with a dash, followed by a speech tag, and then the speech continues without any typograph- ical …

Content truncated to 500 words