Annotating Attribution in Czech News Server Articles

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 1817–1823 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 1817 Annotating Attribution in Czech News Server Articles Barbora Hladk´a1, Jiˇr´ı M´ırovsk´y1, Maty´aˇs Kopp1, V´aclav Moravec2 1Charles University, Faculty of Mathematics and Physics, Institute of Formal and Applied Linguistics 2Charles University, Faculty of Social Sciences, Institute of Communication Studies and Journalism {hladka, mirovsky, kopp, }@ufal.mff.cuni.cz, vaclav.moravec@fsv.cuni.cz Abstract This paper focuses on detection of sources in the Czech articles published on a news server of Czech public radio. In particular, we search for attribution in sentences and we recognize attributed sources and their sentence context (signals). We organized a crowdsourcing annotation task that resulted in a data set of 2,167 stories with manually recognized signals and sources. In addition, the sources were classified into the classes of named and unnamed sources. Keywords: media bias, news server, article, attribution, signal, source, annotation, crowdsourcing 1. Introduction Automated journalism refers to the use of algorithms to automatically generate news from structured data, see e.g., Lepp¨anen et al. (2017). Artificial Intelli- gence (AI) journalism is a broader concept that in- cludes not only automation but machine learning and data processing of various newsrooms related tasks as well. In his survey, Becket (2019) reports that news gathering, news production, and news distribution are the three most common areas for his respondents’ fu- ture AI-tool wishlist. Marconi (2020) provides readers with a more detailed discussion on how journalism will change through the AI-based processes. No doubt Nat- ural Language Processing (NLP) plays (and will play) a key role in AI journalism. One of the areas that is currently receiving a lot of attention is the area of a systematic, empirically- based, and historical-comparative understanding of media bias. Wikipedia defines media bias as “the per- ceived bias of journalists and news producers within mass media in the selection of events and stories that are reported, and how they are covered.”1 Quite sur- prisingly, there are still only few NLP studies system- atically analyzing media bias, see e.g., Hamborg et al. (2019). Currently, significant attention is being paid to fact-checking, see e.g., Thorne and Vlachos (2018), Lazarski et al. (2021). In social sciences, the news production process is an established model that defines nine forms of media bias and describes where these forms originate from, see e.g., Baker (1996), Park et al. (2009). An informative, balanced article should provide the background of a story, including naming sources. This paper aims at a news server of Czech public radio and focuses on detection of sources in its articles. In par- ticular, we search for attribution in sentences and we extract attributed sources from them. This is a task from the area of text understanding and, at least to our 1https://en.wikipedia.org/wiki/Media_ bias knowledge, an NLP system automatically processing attribution in newspapers articles has not been imple- mented yet. Prasad et al. (2006) have proposed and described an annotation scheme for marking the attri- bution …

Content truncated to 500 words