Metadata Games: Improving the discoverability of images by crowdsourcing descriptive tags

Scott Renton, MIMS Project Officer, Digital Library

The University of Edinburgh Library’s media repository (https://images.is.ed.ac.uk) is full of high-resolution digital images of the institution’s unique and valuable collections. Content flows in from BAU digitisation, projects, and readers’ orders and requests, amounting to (currently) nearly 300,000 images available using deep zoom through LunaImaging software, and the IIIF framework.

The nature of digital content’s provenance means that often the metadata describing images is scant; it’s reliant on readers, photographers or the project team providing subjects which describe an image. This burden leads to poor description, which results in poor search within the system.

In a bid to tackle this, the Digital Library’s Development and Systems Team and Digital Imaging Unit collaborated to develop a games-based crowdsourcing application, entitled Metadata Games, enticing participants to tag randomly-generated images from the collections. Incentives, such as coffee vouchers and a real-time leader board, resulted in competitive tagging, and resultantly thousands of tags were harvested and pushed back into the system.

The team took the game to Library sites across the campus, developing two more instances of the game, one based on the Art Collection, with real-live artefacts on show at the sessions, and one based on the Towards Dolly genetics exhibition, entitled “Where’s Dolly?”. The team also partnered with Tiltfactor at Dartmouth College, who had developed a number of their own games; they took some of our content and pushed it into their games, allowing exposure to a new market.

Questions were raised around acceptance and veracity of tags, to address this validation of other people’s tags was an aspect of the game and resulted in points for the participants. Tags would only reach an accepted threshold for ingest if they were validated by other participants. It was also important to carefully mark on the resulting metadata that these tags were citizen- not academically-sourced. Overall, it was a worthwhile activity: fun and innovative, and it made our collections that bit more usable.