Users pick their translations based on several factors:
- Language availability. It is available in a language they can translate from, and it is unavailable in a language they can translate to. The current suggestion system makes sure only valid suggestions in this regard are provided.
- Content quality. The source article has great content (text, images, references, links, etc.) to be reused in the translation. A one line stub is not very useful as a source article.
- Impact. The article is relevant for the target audience, expecting it to be accessed.
- Interest. Topics the user finds interesting and knows or wants to learn about.
When providing suggestions it may be helpful to surface information related to the aspects above. We considered surfacing the following:
- Featured status of the article (quality). This makes only sense when suggestions are not only based on featured articles (T111028) otherwise all suggestions would be highlighted.
- Articles with a high number of views. Articles that are read by lots of people have the potential to be interesting to a lot of people in the target language. In the example we use "10K views" but we may need to adjust the threshold in a way that better convey the idea without resulting in all or any articles being highlighted.
Other options for later consideration:
- Surface other quality assessments such as "good article".
- Some metric based on available content (number of sections, images, links and reference).
- Red links in the target language that match the article. It indicates high demand but the correspondence between concepts may fail.
- Surface relevant article categories the user has previously translated about.