Commons upload script test

[baskaufs]

The test script for uploading Commons items is commonsbot.ipynb. It's current status is that it can upload individual items or a series of items using a loop. The loop uses the idiosyncratic data source upload_metadata.csv, which was purpose-built for uploading the Vandycite videos.

There are several things that need to be improved in this script before it is ready to be used in production:

• [x] The loop part of the script is currently using the Commons "information" template. For the gallery images, it would be more appropriate to use the "artwork" template if information can easily be acquired from the existing gallery metadata. However, we intend to primarily maintain these metadata from the Wikidata item, since it's structured data there and not templated HTML text. So I don't think it's worth taking extraordinary measures to supply those data here.
• [x] The function that adds structured data, create_commons_claim(), creates individual claims one at a time. If there are going to be more than one or two claims added per image, this should be changed from using the wbcreateclaim method to using wbeditentity that VanderBot uses. That method also allows the image Wikibase label (the "image caption") to be set at the same time instead of using the wbsetlabel method used in the set_commons_image_caption() function.
• [x] Related to the previous issue, we should determine what structured data beyond the minimum licensing data should be provided. We learned from the Vandycite video test that media items get flagged if they don't at least have those structured data.
• [x] The import data format needs to be changed from the idiosyncratic format to something more standardized. Andrew Lih (@fuzheado) was creating a YAML mapping template to match API fields with Wikibase properties and it may be possible to use something like that in combination with a CSV to do the mapping. I've pinged him (2021-11-30) to ask him where that code is.

[baskaufs]

I found the example YAML file from Andrew. It's in this gist.

[baskaufs]

I found my notes in a Word doc (action-items.docx): Notes from the 2021-05-04 meeting on Structured Data in Commons (LD4 Wikidata meeting)

Call notes: https://docs.google.com/document/d/1Es7BP60JMv07jZGDmF4TJld7KxZu3cDYYQdzoM_jrRM/edit

Presenters: Alicia Fagerving alicia.fagerving@wikimedia.se John Cummings Wikimedian in Residence, UNESCO

Wikidata+Commons example https://meta.wikimedia.org/wiki/Structured_data_for_GLAM-Wiki/Roundtripping/KMB

Best practice is to put data for the images on Commons and for the things in the image in Wikidata.

o Structured data lets you add data about photo using properties from Wikidata  Depicts--most important property in SDC • What is in the file • Links to Wikidata item • To see all depicts for an item: Type in search field: o haswbstatement:P180=Q750444  Creator • For Wikimedia users: o Object has role: photographer o Author name string: o Wikimedia username o URL:  Inception • Date photo taken  Coordinates • Fetched from exif data

Search engines depend on depicts data, so important.

How to link image of work on Commons side--use "depicts" for what’s in image and use "digital representation of" property to link to Wikidata item

Structured data modeling discussion: https://commons.wikimedia.org/wiki/Commons:Structured_data/Modeling

[baskaufs]

At an LD4 meeting, structured data was discussed. Here are some notes I took:

That last URL leads to the Image Annotator tool: https://image-annotator.toolforge.org/

[baskaufs]

Discovered that there is a Commons category for Vanderbilt University Fine Arts Gallery. We should use it to tag the images when they are uploaded to Commons. Also, we should make sure that the images that are there (mostly Vanderbilt family portraits) are represented as Wikidata items that are linked to the Gallery.

[baskaufs]

For public domain works using the Artwork template, the license property doesn't need to be provided in Structured data. However, it will get flagged if it doesn't have a P6243 (digital representation of) statement in Structured data linking the work back to the Wikidata artwork item. Not sure if that flag gets set when the Wikidata Q ID is given in the template or if it always gets flagged.

[baskaufs]

Incremental work to switch commonsbot script to use artwork template https://github.com/HeardLibrary/linked-data/commit/3a2e1916605ddd3d8dbbf15dc36897ab4f534312

[baskaufs]

Completed with https://github.com/HeardLibrary/linked-data/commit/ce34ff34438db5e0abc87dc769e646d8f8e75cbd