Continuously improve coverage based on usage patterns.
OpenStreetMap provides 710,000+ features tagged with a Wikidata ID. Linking map features to their corresponding Wikidata knowledge gives us access to much more information, including languages. We use these links to source new translations for important features and contribute these translations back into Wikidata. Leveraging Wikidata as a verifiable resource also retains cross-team consistency, meaning both our cartography and geocoding teams can rely on the same data source.
Whether something is important enough to be translated can be derived from different OpenStreetMap tags, i.e. key:population, number of translated name tags, or spatial attributes like covered area. We also match OpenStreetMap with Natural Earth data to find even more labels worth translating.
Adding data to vector tiles means our clusters will regenerate more than 700 million map tiles. We always ensure the new batch of data doesn’t negatively impact rendering, performance, or latency of our users.
To add language fields to our vector tiles, we extend our PostGIS based rendering stack with queries for
additional name fields in our localized label layers.
We also look for common translation quirks such as single/double quotes, apostrophes, text in parentheses, articles, and other
problematic patterns in labels we’ve seen over the years.
Expect to see many more internationalization features in the coming months. We plan to make our maps more globally accessible, allowing developers building with Mapbox greater reach into fast-growing markets. Stay tuned!
Join us and push maps to the next level
The cartography team is on the forefront of digital mapping. We push the limits in leveraging and curation of open source data in hopes to build the largest spatial map of human knowledge ever created through our processing pipeline.