MaSS: A Large and Clean Multilingual Corpus of Sentence-aligned Spoken Utterances Extracted from the Bible

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 55–64 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 55 GerDraCor-Coref: A Coreference Corpus for Dramatic Texts in German Janis Pagel, Nils Reiter University of Stuttgart, Institute for Natural Language Processing (IMS) Pfaffenwaldring 5B, 70569 Stuttgart, Germany {firstname.lastname}@ims.uni-stuttgart.de Abstract Dramatic texts are a highly structured literary text type. Their quantitative analysis so far has relied on analysing structural properties (e.g., in the form of networks). Resolving coreferences is crucial for an analysis of the content of the character speech, but developing automatic coreference resolution (CR) systems depends on the existence of annotated corpora. In this paper, we present an annotated corpus of German dramatic texts, a preliminary analysis of the corpus as well as some baseline experiments on automatic CR. The analysis shows that with respect to the reference structure, dramatic texts are very different from news texts, but more similar to other dialogical text types such as interviews. Baseline experiments show a performance of 28.8 CoNLL score achieved by the rule-based CR system CorZu. In the future, we plan to integrate the (partial) information given in the dramatis personae into the CR model. Keywords: corpus, coreference, computational literary studies, literary text, dramatic text 1. Introduction This paper introduces a new annotated corpus. A selection of (parts of) German dramatic texts (1730-1920 CE) has been annotated manually with coreference information, to be used as a gold standard for experiments on automatic coreference resolution (CR). Coreference occurs when en- tities are referenced multiple times in a text, to be seen in Figure 1: All occurrences of the name Romeo refer to the same character, and the same is true for Juliet. Addition- ally, the phrases a window, yonder window and It refer to the same window; the sun and fair sun refer to the sun. Next to prose and poetry, drama is one of the three major literary genres, and an interesting text type with respect to both its linguistic and literary properties. Dramatic texts are highly structured: Speakers and stage directions are clearly marked (by typography or in machine-readable XML, see Fig. 1 for an example) and separated from the text charac- ters utter. Spoken text is (most of the time) clearly associ- ated with the character who utters it, and the entire text is hierarchically segmented into acts, scenes, and, sometimes, appearances. While dramatic texts can be considered as the ‘template’ for theatrical productions, not all dramatic texts have been written with the intention of being produced on stage, and not all theatre plays are based on dramatic texts. Due to the strong structure, and the fact that this structure has been encoded in TEI/XML frequently, it is straightfor- ward to access some properties of the texts, without any language analysis per se. For instance, extracting character networks that are based on co-presence is a major research trend within the computational literary studies community (e.g., Trilcke et al. (2015)). This can …

Content truncated to 500 words