CATAMARAN: A Cross-lingual Long Text Abstractive Summarization Dataset

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 707–714 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 707 Subjective Text Complexity Assessment for German Laura Seiffe1, Fares Kallel1, Babak Naderi2, Sebastian M¨oller1,2, Roland Roller1 1German Research Center for Artificial Intelligence (DFKI), Speech and Language Technology Lab, Berlin, Germany firstname.secondname@dfki.de 2Quality and Usability Lab, TU Berlin, Berlin, Germany firstname.secondname@tu-berlin.de Abstract For different reasons, text can be difficult to read and understand for many people, especially if the text’s language is too complex. In order to provide suitable text for the target audience, it is necessary to measure its complexity. In this paper we describe subjective experiments to assess the readability of German text. We compile a new corpus of sentences provided by a German IT service provider. The sentences are annotated with the subjective complexity ratings by two groups of participants, namely experts and non-experts for that text domain. We then extract an extensive set of linguistically motivated features that are supposedly interacting with com- plexity perception. We show that a linear regression model with a subset of these features can be a very good predictor of text complexity. Keywords: Corpus Creation, Readability Assessment, Complexity 1. Introduction The ability to read and to understand text is a crucial com- petence to communicate and to exchange information. For different reasons, understanding written language can be a challenge not only for different target groups like language learners, students, or people with cognitive limitations, but also for people with less experience or knowledge about the text’s content. All these user groups have different needs in terms of readable texts. To enable people to understand text is not only important for general communication, but also plays a long-term role for inclusion and integration. Addi- tionally, user-friendly and user-centring applications need text that is well adapted to the needs of their target groups. Besides understanding which texts are hard to read for spe- cific target groups, the assessment of readability is an im- portant intermediate target for text simplification - knowl- edge about the obstacles the different reader groups have, is crucial to make text easier. A machine based evaluation of text readability can be used to get insights into this quality aspect of text. The definition of the term readability as well as the closely related terms understandability and (text) complexity are widely discussed in the field of research (Zamanian and Heydari, 2012). Linguists, educationalists, psychologists, and machine learning engineers have different perspectives on that. Whether the text is human made or machine gen- erated plays a role for the definition of readability as well (Howcroft et al., 2020). Generally, readability is often very vaguely defined as how easily a written text is to read. In other cases, readability is equated with linguistic complex- ity. Understandability usually refers to the understanding of the text and examines the reader’s perspective: The un- derstanding of a text depends on their prior knowledge of topic and language (Mclaughlin, …

Content truncated to 500 words