The_Tower_of_Babel.jpg: Diversity of Visual Encyclopedic Knowledge Across Wikipedia Language Editions

Abstract

Across all Wikipedia language editions, millions of images augment text in critical ways. This visual encyclopedic knowledge is an important form of wikiwork for editors, a critical part of reader experience, an emerging resource for machine learning, and a lens into cultural differences. However, Wikipedia research--and cross-language edition Wikipedia research in particular--has thus far been limited to text. In this paper, we assess the diversity of visual encyclopedic knowledge across 25 language editions and compare our findings to those reported for textual content. Unlike text, translation in images is largely unnecessary. Additionally, the Wikimedia Foundation, through the Wikipedia Commons, has taken steps to simplify cross-language image sharing. While we may expect that these factors would reduce image diversity, we find that cross-language image diversity rivals, and often exceeds, that found in text. We find that diversity varies between language pairs and content types, but that many images are unique to different language editions. Our findings have implications for readers (in what imagery they see), for editors (in deciding what images to use), for researchers (who study cultural variations), and for machine learning developers (who use Wikipedia for training models).

Disappearing without a Trace: Coverage, Community, Quality, and Temporal Dynamics of Wikipedia Articles on Endangered Brazilian Indigenous Languages

“…Given its popularity and accessibility, many researchers have conducted studies in Wikipedia, exploring various aspects of the platform, such as its content, structure, and social dynamics. Those studies have provided insights into issues such as automatically assessing article quality (Dalip et al 2009;Shen, Qi, and Baldwin 2017), analyzing citations (Piccardi et al 2020;Baigutanova et al 2023), images (He et al 2018;Rama et al 2022), andinfo-boxes (Graells-Garrido, Lalmas, andMenczer 2015;Lewoniewski 2017) or understanding readers' preferences (Lehmann et al 2014). Other works focused on identifying underrepresented groups (Graells-Garrido, Lalmas, and Menczer 2015;Mandiberg 2023;Gallert et al 2016;Sethuraman, Grinter, and Zegura 2020;Hoenen and Rahn 2021) and the asymmetry of the coverage across different language versions (Hale 2015;Graham 2011;Lemmerich et al 2019;Roy, Bhatia, and Jain 2020;Ashrafimoghari 2023).…”

Characterising User Content on a Multi-Lingual Social Network

“…For example, language differences amongst Wikipedia editors (Kaffee and Simperl 2018), as well as general differences amongst language editions (Hale 2014;Hecht and Gergle 2010;Hale 2015). Often this goes beyond textual differences, showing that even images differ amongst language editions (He et al 2018). There have also been efforts to bridge these gaps (Bao et al 2012), although often differences extend beyond language to include differences between facts and foci.…”