Opportunities for symbiotic and effective use of automated metadata tagging.
The labour required to maintain high-quality metadata across all flavours of content technology — from traditional document management to the more modern marketing technology platforms — can often seem sisyphean: such that it can never be completed. Like Sisyphus, we roll the same big rock up the hill every day, manually adding metadata in a valiant attempt to achieve new levels of automation and smarter business outcomes. Still, as the scope of what we do with Digital Asset Management (DAM) technology grows, the current ways of maintaining metadata and quality assurance become increasingly unsustainable.
About a decade ago we began to hear about the promise of auto-tagging or auto-recognition technology, whereby systems would ingest an image and be able to identify not only basic information like colours and objects, but also people, brand logos, and even specific places or a person’s ethnicity. But much like speech recognition technology, initial forays into auto-tagging in commercial content management systems were fraught with errors, to the point where it made more mistakes than it did accurate identifications.
But now, that’s changing.
What is Auto-Tagging?
Specifically in the context of Digital Asset Management, auto-tagging refers to automatically generating metadata in order to catalogue assets. The image is scanned, and patterns are identified. In some cases, embedded metadata from a camera or other content creation system might also be read or identified, and then turned into metadata within the DAM. Over the life of the asset, such automated tags may accumulate to include not just what’s in the asset, but also how it’s used, where it’s used, who’s viewing it, and their relationships to each other.
Calling auto-tagging “artificial intelligence” is (in my opinion) premature, but it sounds fashionable and modern, so you will often hear technology vendors confounding the terms. Let’s be honest: auto-tagging technology isn’t going to pass the Turing test anytime soon. At the core, it’s not always that intelligent – it’s really just identifying patterns (albeit complex ones at times), or mining already-existing data in moderately smart ways. What remains to be seen is whether the big leap will happen: if these algorithms can improve themselves, using machine learning to become smarter.
What is the promise?
The much-hyped promise of auto-tagging technology is that these solutions will replace the manual work and effort (or allow work and effort to be re-directed) that’s currently undertaken by brand managers, agencies, librarians, and data migration teams to describe and classify both assets and content. There’s much expounded about how these technologies will reduce time and cost while ensuring enriched metadata, all the while improving accuracy and consistency, and improving search results. Subsequently you’ll realise more user adoption, experience less user frustration and push-back, and increased ROI.
Not so fast! You shouldn’t be jumping to outsource all your metadata tagging to a robot. The reality is, despite the advancements in automated tagging during the last two years in particular, it’s still no replacement for subject matter expertise, customer experience awareness, and brand knowledge. Vendors, content creators, brands, and service providers are on a long journey that requires a symbiotic relationship between these new technologies and people who are knowledgeable about the desired customer experience outcomes.
Proof of Concept with ICP
I recently collaborated with the team at ICP to write a detailed white paper about their auto-tagging proof of concept, whereby they tested various auto-tagging tools against a corpus of assets from a major multi-brand enterprise. For a company like ICP, the time was ripe to explore auto-tagging technology and uncover what benefits it could bring. The paper explores the opportunity of auto-tagging in the context of digital asset management technology, and reports on ICP’s testing of various offerings.
You can also contact me via LinkedIn or my web site, if you have any questions about this ever-changing technology landscape.