Supporting multilingual editors in small Wikipedias who are leveraging translation to contribute across knowledge and content gaps.
Worldwide, often countries with the fastest growing online populations are home to languages with relatively small Wikipedias (wikis). Thus, if the Wikimedia Foundation will achieve its goal of removing barriers to knowledge, it must address language barriers and the socio-cultural needs of these communities, including many in South and Southeast Asia.
Despite being relatively “small”, many wikis in these regions have significant growth potential based on language community size and number of active editors. Adding to this potential of increasing knowledge access, previous research has indicated that multilingual editors may be more active than their ‘single edition (monolingual)’ counterparts. This suggests that these editors are important actors for helping to close content gaps, or areas of knowledge that are only accessible on certain language wikis but not others.
An important way of addressing language barriers and content gaps is through translation. As of December 2020, the Wikimedia Foundation’s Language Team’s Content Translation tool has been used to create over 750,000 articles since its inception.
As part of its annual plan, the Language Team aims to grow small wikis with translation and improve the process of creating multilingual content. In order to reach the goal of helping small wikis grow with translation, the team aims to better understand the needs of small wikis and editors in these communities. This includes identifying barriers and challenges to the use of translation and creation of new content. To help make multilingual editors more productive in creating new content, this generative research project aimed to investigate the current experiences of current editors, and better understand workflows of potential editors who currently create online content with the aid of translation. For both current and potential small wiki contributors, the current project focused on workflows, Wikipedia experience journeys, motivations, needs, and barriers to editing.
The primary overarching research goal of this generative research project was to better understand multilingual editor experiences in small wikis, particularly those involving translation. More specifically, we wanted to increase our knowledge of small wiki multilingual editor experiences - both current and potential contributors. This involved investigating the socio-cultural context of the target communities, and also identifying specific language-related considerations and concerns. These language factors included structural, written, and sociolinguistic factors. Key sub areas of focus included editor values and motivations, barriers to content creation, and awareness/knowledge of the Wikipedia ecosystem, especially translation and multilingual features. Using phenomenological methods, the project emphasized task-based inquiry to uncover content creation workflows and provided evaluative assessment of current and potential contributor experiences with current Wikipedia translation tools. Research planning also involved a desk review of relevant previous research on translation and editing. An annotated bibliography titled 'Multilingual Editors and Translation' was prepared to summarize a selection of relevant articles and readings.
- How do small wiki contributors edit and create content? What are typical content creation and translation workflows? To what degree (and how) are these workflows exclusively supported by Wikipedia, or not?
- How do editors discover and experience Content Translation? What are common entry points, discoverability problems, usability issues, and barriers/motivations to use?
- What are multilingual editors’ motivations to use translation, and what Content Translation value propositions resonate most?
- For design purposes, what are small wiki editor personas?
- What barriers do editors face when using Content Translation? Why do some continue to use it while others stop?
- How do users understand and think about translation, especially as it relates to Wikipedia, machine translation (MT), and use of MT in tools?
Please refer to the full report, where you can find more details on the project and research approach, as well as detailed results, opportunities, personas, and experience journeys.
10 Key Takeaways
- Most participants, including both current and potential editors, were unable to successfully navigate to Content Translation (CX), despite consistently searching in many of the same places. Many pre-translation workflows begin at large wikis where editors are reading.
- Current editors and newcomers are delighted by a number of CX features that speed up the editing process and require less typing, a challenge for some users. Delighters included automatic links and references, side-by-side presentation of the source/target article, and help generating key article vocabulary.
- Keyboard input challenges with non-Latin scripts are widespread. This is not a CX-specific problem, but impacts the CX experience in critical ways. While some current editors have mastered non-Latin inputs on WP, others have developed a range of workaround solutions. Without the ability to type fluently, users are unable to edit machine translation outputs.
- Observation of editing workflows revealed that 17 of 21 current editors sampled relied on external tools for at least one of the following purposes: input, drafting, translation. These external supports are helping editors accomplish goals such as entering non-Latin scripts, drafting new articles, and translating article sections.
- Alerts/warnings in Content Translation are not always well-understood and timing of their presentation is critical. Editors didn’t always understand how to to resolve alerts, and are not clear on how to proceed if ready to publish but not able to due to a machine translation use limits.
- An editor’s motivation is multi-dimensional. Motivations range from purely altruistic in nature to those grounded in self-improvement. During research sessions, we observed 6 primary patterns: improving access, expanding contents, self-improvement, social, audience, recognition/identity (each detailed in full report).
- Small wikis present unique environments for editors compared to larger wikis. For example, editors noted that when compared to the English wiki, there’s often much less anonymity and more receptiveness to new editors. On the other hand there are fewer editors to complete admin tasks, proofread articles, and help find sources.
- In most cases, participants faced problems related to a specific task. They wanted to troubleshoot a very specific, contextual problem, not access a lengthy help page. New translators may also benefit from the support systems that experienced translators report having; for example, knowing who other active translators on their wiki are, and who to approach for translation support.
- Small wiki editors vary in terms of whether they are primarily mobile-first editors or rely on laptops/chromebooks. In general, editing and reading activities correlate with device use patterns. In general, we found that input of non-Latin scripts is slightly easier for potential editors on mobile devices.
- Despite nearly 100% brand awareness, the editing process remains below the tip of the Wikipedia iceberg. Multilingual readers often default to reading on larger wikis, and translation as a form of editing lacks widespread awareness - even among current editors. Social factors are in part determining how editors contribute and what topics they contribute about.
Design deliverables: Personas and Experience Journeys
- To explore these trends, you can access any source of online population trends and compare with current language Wikipedia sizes.
- For strategy purposes, in their annual plan, the WMF Language Team has defined "small" as wikis with fewer than 100k artices and less than 100 translations per month. The threshold of 70 editors (and >20 active editors) per month is hypothesized as important figures for potential growth.
- See Scott A. Hale's (2014) paper titled 'Multilinguals and Wikipedia Editing'
- Content Translation Statistics