Identification of Biased Terms in News Articles by Comparison of Outlet-specific Word Embeddings Timo Spinde 1,2[0000-0003-3471-4127] , Lada Rudnitckaia 1[0000-0003-2444-8056], Felix Hamborg 1 [0000-0003-2444-8056] , and Bela Gipp 2[0000-0001-6522-3019] 1 University of Konstanz, Konstanz, Germany {firstname.lastname}@uni-konstanz.de 2 University of Wuppertal, Wuppertal, Germany {last}@uni-wuppertal.de Abstract. Slanted news coverage, also called media bias, can heavily influence how news consumers interpret and react to the news. To automatically identify biased language, we present an exploratory approach that compares the context of related words. We train two word embedding models, one on texts of left- wing, the other on right-wing news outlets. Our hypothesis is that a word's rep- resentations in both word embedding spaces are more similar for non-biased words than biased words. The underlying idea is that the context of biased words in different news outlets varies more strongly than the one of non-biased words, since the perception of a word as being biased differs depending on its context. While we do not find statistical significance to accept the hypothesis, the results show the effectiveness of the approach. For example, after a linear mapping of both word embeddings spaces, 31% of the words with the largest distances po- tentially induce bias. To improve the results, we find that the dataset needs to be significantly larger, and we derive further methodology as future research direc- tion. To our knowledge, this paper presents the first in-depth look at the context of bias words measured by word embeddings. Keywords: Media bias, news slant, context analysis, word embeddings 1 Introduction News coverage is not just the communication of facts; it puts facts into context and transports specific opinions. The way how "the news cover a topic or issue can deci- sively impact public debates and affect our collective decision making" [12], slanted news can heavily influence the public opinion [11]. However, only a few research projects yet focus on automated methods to identify such bias. One of the reasons that make the creation of automated methods more difficult is the complexity of the problem: How we perceive bias is not only dependent on the word itself, but also its context, the medium, and the background of every reader. While many current research projects focus on collecting linguistic features to describe media bias, we present an implicit approach to the issue. The main question we want to answer is: T. Spinde, L. Rudnitckaia, F. Hamborg, B.Gipp “Identification of Biased Terms in News Articles by Comparison of Outlet-specific Word Embeddings”. In: Proceedings of the iConference 2021. DOI: 10.1007/978-3-030-71305-8_17 Preprint from: https://www.gipp.com/pub/