GapFinder/Translation
 | This page is obsolete. It is being retained for archival purposes. It may document extensions or features that are obsolete and/or no longer supported. Do not rely on the information here being up-to-date. This extension will be deprecated by 15 July 2024. If you have code that relies on the API, you should migrate it to LiftWing (example). Users of the UI should instead use Content translation. |
| GapFinder |
|---|
| Try the tool |
| Gap Types and usage |
| Feedback |
Recommend articles for creation that exist in one language but are missing in another.
Usage
[edit]Get recommendations
- Pick a source language you want to translate from
- Pick a target language you want to translate to
- Optionally, input an article's title into the search bar and press Enter to return results related to that article
Begin translating
- Click on a recommendation to open a preview of the existing article
- Either
- Click Create from scratch to begin creating the article from scratch in the target language
- Click Translate to begin translation in the Content translation tool
Features
[edit]We know from the first experiment that personalized recommendations perform much better than random recommendations when it comes to user activation rates and article creation rate.[1] Ideally, we want to have the personalized recommendations based on users' edit history available in the tool. However, we also want to reduce the barrier for entry by showing recommendations to users before they are required to login via centralAuth, and we also want to allow users with almost no edit history to be able to get recommendations about articles they may be interested in. We make this possible by providing an option for the user to enter an article that they would have liked to edit (or they have edited and liked it) in the source language of their choice, and get recommendations that are similar to their article of liking in the destination language based on a topical model that algorithm operates on.
How does it work?
[edit]The recommendation works in two ways. Once the source and destination languages are chosen, a random subset of articles that are available in the source language but are missing in the destination language and are predicted by the algorithm to be eligible to exist in the destination language are shown to the user. The user can receive personalized recommendations by providing an article name in the source language that the user would contribute to in the destination language and by clicking Recommend, the user receives similar articles missing in the destination language but available in source.