Feature summary:
On the participate page, we need to show depicts suggestions from the Metadata-to-Concept Module (currently in development, see sample requests and responses here).
These should display alongside the current Google Vision suggestions (only visible when the ?mv=true parameter is added to end of the URL).
They will work in the same way. Simply click on a suggestion to add it to the list of depicts changes. The user still needs to click save to actually add the statement.
See design here: T312873: Create mock designs for depicts suggestions from Google Vision and Metadata-to-Concept
Update 7 Oct 2022
a first prototype of the metadata-to-concept API is now up and running on a virtual server listening on http://92.205.161.71:8080/extract/
It now has a basic concurrency model and should (in theory) be able to handle multiple requests simultaneously.
The attached screenshot should be self-explanatory, nevertheless here are a few notes:
- The keys in the first level JSON dict are: name, url, attrs. All of them are mandatory. The name and url keys must be strings, the attrs key must be a list of dicts.
- At the moment name and url are ignored.
- All attributes used to derive concepts must be declared in the attrs list. Each attribute has a key, a value and lang. Lang is currently ignored and defaults to "en". One should always add the attribute name (= file name) as a fallback in the list.
- At this point the attribute types name, description, caption and categories are processed. Categories must be a semicolon-separated list of wiki categories.