For users that want to translate but they don't have a clear idea on what to translate, it would be helpful to show some suggested articles.
Presenting suggestions
Suggestions will be presented as a new view (along with ongoing and published translations) and will allow users to discard or make them stick in addition to allow users to start translating.
General goals
@Nirzar has been working on some designs for this. The main goals for the design are:
- Make suggestions more prominent when those are more needed (e.g., suggestions can take the main translation list area when it is empty).
- Avoid an up-front configuration process (users should be able to get suggestions immediately).
- Don't block the user progress (i.e., suggestions not getting in the way)
- Encourage users to keep translating.
The suggestion list
Suggestions will be presented as a new type of list provided at the translation dashboard.
- The listing will be accessible as part of the ame selector that allows to switch between in-progress and published translations.
- The suggestions view will be the default view when the user has no in-progress translations. If there are in-progress translations, the in-progress view will be the default one once the user reaches the translation dashboard.
- When the user lands in the in-progress view a small indicator of new suggestions will be shown to communicate that there are new suggestions. 3 new suggestions can be announced when the user access the dashboard for the first time on a given day.
- The suggestion list will be affected by the general language selectors:
- Initially the language selector will be set to the default value (from "all languages" to "all languages"). For that state we only want to show suggestions for language pairs we are very confident the user speaks (not random ones). So we may want to show suggestions from languages we are very sure the user speaks.
- If a language pair is selected, the list of suggestions will show only suggestions for that language pair.
- Each suggestions will display enough information to inform the user decision on whether to translate it or not:
- Article title in the source language.
- Wikidata description in the current UI language (if available) or the source language.
- Article image.
- Languages. The source and target languages if the language selector is in the default state. We may consider showing the languages (from those we are sure the user knows) the article is missing (e.g., Missing in Catalan, French and 5 more").
- Quality/Impact assessments. Highlighting articles that are featured may encourage users to translate them.
Showing new suggestions
- Initially 15 suggestions will be shown.
- Every day the user access the dashboard, new suggestions are added.
- New suggestions will be highlighted to differentiate them from the rest. Note that suggestions are incrementally presented to the users to make them easier to digest even if the underlying service provides us with a single long list.
- Users can request more suggestions explicitly.
- New suggestions will appear on top pushing the existing suggestions down.
- Very old suggestions can be removed automatically to avoid exposing the users through the same suggestions agai and again.
Operations on suggestions
- Clicking on a suggestion opens the new translation dialog with all the information pre-filled.
- Users can discard a suggestion. That will remove it from the list and a new suggestion will appear. In the future, a follow-up process can be defined to guide the kind of suggestions to show.
- Users can mark a suggestion for later. That is low-commitment step that puts a suggestion closer to become a translation in the near future (view prototype).
- When a suggestion is marked, it will appear on top with a more compact styling and a clear call to action to start translating.
For the future
When completing the designs we may want to make it possible to integrate our solution with another related feature: user-made list of articles to translate later. That has not been discussed yet, but we may want to explore if user-generated and machine-suggested articles can be presented to the user in a consistent way.
Obtaining the suggestions
The ultimate goal is to provide suggestions to the user about articles that (a) are interesting to the user, (b) the source article has great content to translate (e.g., suggest to translate a one-paragraph stub may not be the best choice), and (c) it has impact (e.g., the source article is viewed a lot).
Finding good articles to suggest seems a complex problem, so I think we can start with a simple algorithm and keep improving it based on measuring how many people starts articles using such suggestions. Some ideas for simple approaches:
- Articles only available in a language.
- Featured articles missing in a language.
- Common red links users find while translating.
- Most accessed missing pages.
@leila had some very interesting ideas on more sophisticated mechanisms to identify relevant missing articles, so any progress in that area can be used to improve our translation suggestions.