Multilingual Sentence-Level Bias Detection inWikipedia

Abstract

We propose a multilingual method for the extraction of biased sentences from Wikipedia, and use it to create corpora in Bulgarian, French and English. Sifting through the revision history of the articles that at some point had been considered biased and later corrected, we retrieve the last tagged and the first untagged revisions as the before/after snapshots of what was deemed a violation of Wikipedia's neutral point of view policy. We extract the sentences that were removed or rewritten in that edit. The approach yields sufficient data even in the case of relatively small Wikipedias, such as the Bulgarian one, where 62k articles produced 5k biased sentences. We evaluate our method by manually annotating 520 sentences for Bulgarian and French, and 744 for English. We assess the level of noise and analyze its sources. Finally, we exploit the data with well-known classification methods to detect biased sentences. Code and datasets are hosted at https://github.com/ crim-ca/wiki-bias.

Examining Linguistic Biases in Telegram with a Game Theoretic Analysis

“…Our model, however, sharpens the notions of framing and coverage biases by linking them to strategies at the lexical and discursive level that can be opportunistic and evolve over time. We also see the confirmation of different levels of granularity in our corpus: category-level, message-level and media source level [1,7].…”

“…4.3, questions lexicon and embedding-based approaches. In addition, seemingly neutral labels (such as people vs. ordinary people vs. people in plainclothes) are usually not considered as potentially biased, especially in lexicon-based approaches, such as [1].…”

“…To see some examples of strategies, let's return to (1). Suppose 0 has played ρ 1 ; one strategy of 1 is to play a clarification question ρ 2 like did you mean that the shouts of "Shame" were addressed to the court?…”

“…1. Dynamics of the conversation extended conversation as an ME game G between the protesters (0) and the government (1). Player i constructs a narrative during his turn either about some contemporaneous event or as a reply and "counter-narrative" to player 1 − i's narrative on some previous turn.…”

Detection of Puffery on the English Wikipedia

“…Prior research on identifying problematic edits has considered weasel words, which Wikipedia defines as "words or phrases aimed at creating an impression that something specific and meaningful has been said, when in fact only a vague or ambiguous claim has been communicated" (Wikipedia, 2021c), and hedges, which Farkas et al (2010) define as phrases "indicating that authors do not or cannot back up their opinions/statements with facts. The best models for detecting weasel words in both the CoNLL-2010 ACL shared task on weasel words (Farkas et al, 2010), and a multilingual weasel word corpus (Aleksandrova et al, 2019) use bag-ofwords classification approaches (Georgescul, 2010;Aleksandrova et al, 2019).…”