Wikidata:WikiProject Wiki Art Depiction Explorer
The Wiki Art Depiction Explorer (WADE) is a user-friendly interface to add depiction information to the Wikidata items of artworks. It was launched in the summer of 2019, and is open for community testing.
The project is a collaboration between Wikimedia District of Columbia and the Smithsonian Institution, with community members Andrew Lih, Robert Fernandez, and Effie Kapsalis working with developer Edward Betts to develop the system. Its development was made possible through support from the Knight Prototype Fund, an initiative of the John S. and James L. Knight Foundation.
Try it out at http://art.wikidata.link
The team conceived of the Wiki Art Depiction Explorer for the Knight Foundation's inaugural Arts and Technology program given their open call for:
"ideas to explore digitally-driven approaches that galleries, museums, performing arts centers, theaters, and arts organizations of all genres might use to inspire audiences."
Wikidata has been on the rise among GLAM/cultural institutions, and adding meaningful depiction metadata for items about artworks has been a main attraction. Our proposal focused on the persistent challenge of making a user-friendly way to "enrich semantic data about artworks in Wikidata, making it free for the world."
Adding depiction statements manually to Wikidata items via property P180 – depicts (P180) – is typically haphazard, unguided, and unassisted. A user has no guidance as to how previous Wikidata items have been used in depiction statements. There is no easy way to work with coherent sets of artworks by the same painter, within an artistic movement, or part of a particular collection. Several questions motivated our thinking:
- How might we visualize the state of depiction metadata for groups of artworks and provide direction for collaborators?
- Could we provide an interface to add meaningful metadata without having to learn Wikidata's intricacies?
- How might a context-driven metadata interface support consistent crowdsourced results for describing art?
- How might we present metadata from institutions to guide volunteers to enrich the Wikidata descriptions?
The Knight Foundation supported this idea and WADE was selected from hundreds of proposals as one of the 12 projects funded that year. (See announcement)
WADE's most basic mode of operation presents the user with a web interface that includes the Wikidata item of an artwork and a large image of the artifact. WADE prompts the user to type in what is depicted in the given painting. We are focusing on paintings for simplicity, though any Wikidata item can be edited using WADE.
Entering terms. As the user types, matches are interactively displayed, and previously used depiction terms listed first. How often that term has been in previous depiction statements is also displayed. Wikidata terms that have never been used before in a depiction statement are listed in a secondary set.
Selecting the name of a matching term adds that to a working queue of terms and any number of depiction statements can be added in this way.
Clicking the "Add these to Wikidata" button commits these changes immediately to Wikidata under the user's account. Therefore, the user needs to have a Wikimedia account. (WADE uses OAuth to authorize the tool to edit on their behalf.)
Pre-existing terms. If there are pre-existing depiction statements for the Wikidata item, they are displayed above the text entry field. There are also additional fields of information related to the artist, collection, and GLAM institution that holds the item.
Navigation. After depiction information has been added to the Wikidata item, a number of options are provided to navigate to similar works by artist, by collection, or by main subject. This provides incentives to keep working on more depictions.
Another option is to start with a master browsing interface to examine artworks by any number of criteria, such as country of origin, fabrication method, artist, etc. The interface is designed to allow users many degrees of freedom to simultaneously explore and discover the artworks in Wikidata while meaningfully editing and adding metadata. Clicking on any of these criteria brings up a gallery of images that users can browse through until a painting captures their attention. Users can further filter by clicking on "toggle filters" to do additional filtering.
The filtering mechanism was meant to be easy to use, with all of the filtering parameters easily readable and customizable in the URL. Therefore, a page describing paintings from the Smithsonian American Art Museum that are landscape paintings would look familiar to those with basic knowledge of Wikidata properties and items:
Fortunately, one doesn't need to know any of these Q and P numbers to use WADE. It can be operated entirely through a point and click interface with no Wikidata experience, which was one of the project's goals.
Sidebar in Wikidata To more conveniently use WADE within Wikidata, you can add a "WADE" link in the sidebar when visiting the page of a Q item by adding this code to your common.js file. When you are on the Wikidata item page for an artwork, click on the WADE link and you will bring up WADE's interface for adding depiction information. The following line can be added to a user's common.js file to enable it.
For some artworks, WADE can import useful context from an institution's object page on web to provide more direction. For example, for Smithsonian American Art Museum (SAAM) paintings, WADE displays descriptive text and a list of keywords that have been gleaned from the museum's page. For a painting like "Two Comanche Girls," WADE will show the text, "the wigwam of the Chief, his dogs, and his five children" and provide controlled vocabulary and subject headings from the museum, such as "Ethnic – Indian – Comanche." Providing this information may help the user to enter depicts statements that they might not have otherwise considered and provides detailed information they can use to add terms to Wikidata.
Another interesting enhancement is related to portraits: if the title of a work contains a start and end date (eg. 1754-1802), WADE will suggest people in Wikidata whose birth and death dates coincide with those years, providing a list to select from. This helps insure the correct person is identified and adds utility to WADE's game-like interface.
Code and reporting
- Source code for the project can be found at the Github repository
- Issues can be reported at the Github issues tab
- Announcement of the tool: in the November 2019 GLAM Newsletter
- Contact: Andrew Lih, project manager
Ideas for the future
We're glad to see the tool has received very positive reviews so far, with thousands of edits made already. Since it is an OAuth tool that adds a well known tag to all edits, it is easy to inspect all changes done via WADE through a special recent changes filter.
Some features that we can imagine exploring include the use of artificial intelligence and machine learning to add to the richness of recommended depiction statements, based on previous work Andrew Lih has done with The Metropolitan Museum of Art, as described in the January 2019 GLAM Newsletter.
As we expand beyond just paintings, how might we work with not just 2D works, but 3D works of art more effectively?
We welcome feedback and more ideas.
During the development of the project, there was feedback from Smithsonian Institution curators and staff and the Wikimedia community, especially those involved with the Sum of All Paintings project.