Identification of Biased Terms in News Articles by Comparison of Outlet-Specific Word Embeddings

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 leftwing, the other on right-wing news outlets. Our hypothesis is that a word's representations 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 potentially induce bias. To improve the results, we find that the dataset needs to be significantly larger, and we derive further methodology as future research direction. To our knowledge, this paper presents the first in-depth look at the context of bias words measured by word embeddings.

What company do words keep? Revisiting the distributional semantics of J.R. Firth & Zellig Harris

“…While variants of this approach are most commonly used to study semantic shifts (e.g., Garg et al, 2018;Kozlowski et al, 2019;Mendelsohn et al, 2020), it could plausibly be used to stratify a dataset according to other variables such as space (e.g. Bamman et al, 2014;Gong et al, 2020), online communities (Lucy and Bamman, 2021), persons (Yao et al, 2020), or domains (Spinde et al, 2021). Words still only keep the company of one another, but by limiting their company we implicitly introduce other participants in the analysis.…”

A domain-adaptive pre-training approach for language bias detection in news

“…Unfortunately, the diversity of online news sources also opens the door for slanted and non-neutral news coverage [36]. Biased news coverage -referred to as media bias in the literature [35,42,43] -occurs once subjective reporting on a specific event replaces objective coverage. Media bias manifests in various forms such as bias by word choice [45] or bias by omission [25] of information.…”

Exploiting Transformer-based Multitask Learning for the Detection of Media Bias in News Articles

“…To date, only a few research projects focus on the detection and aggregation of bias [6,16]. One of the reasons that make the creation of automated methods to detect media bias a complex task is often the subtle nature of media bias, which represents a challenge for quantitative identification methods [10,16,30,33]. While many current research projects focus on collecting linguistic features to describe media bias [11,23,29,34], we propose a Transformer-based [39] architecture for the classification of media bias.…”