Bridging LGBT+ Content Gaps Across Wikipedia Language Editions

Abstract

In the past several years, the Wikimedia Movement has become more aware of the lack of representation of specific communities, that is, content gaps. Next to geographical and gender-related initiatives, the LGBT+ Wikimedia community has organized to create LGBT+ content encompassing (among other topics) biographies, events, and culture. In this paper, we present a computational approach to collecting and analyzing LGBT+ articles. We selected 14 Wikipedia language editions to study the coverage of LGBT+ content in general, its visibility in the list of Featured Articles, and its overlap with the local content of the Wikipedia language editions. Results show that a considerable part of potentially LGBT+ related content exists across Wikipedia language editions; however, this relation is not evident in each language edition. In this sense, closing the LGBT+ content gap is about creating articles and making connection to the topic visible in already existing articles. We also analyze the frequency of biographies of persons with non-heterosexual sexual orientations. We find that even though they represent only a small share of all biographies, they are a bit more frequent among the Featured Articles. When taking into account all the LGBT+ biographies of the different languages, English context celebrities are the most visible. While part of the LGBT+ content is related to each language edition's local context, it tends to be less contextualized than the entire language editions. This indicates the possibility of growing LGBT+ content in each Wikipedia language edition by representing its most immediate LGBT+ local context. We propose a dashboard tool to find relevant LGBT+ articles across language editions and start bridging the gaps. Finally, we conclude this study by presenting recommendations for the next steps amongst the Wikipedia communities to fill some of these gaps.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 9 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.