AI-Friendly Content & Metadata Taxonomy
Utrip
AI-Friendly Content
Problem
We had to build a database for things to do in as many destinations around the world as we could. This not only meant researching, curating, and writing/editing content, it also meant making sure that the content interacted correctly with backend algorithms.
And then there was the scaling issue, to boot.
Solution
I made—and oversaw the creation of—tens of thousands of pieces of content, each of which would surface based on a user’s preferences.
Result
The best AI-based, personalized travel recommendations
Tens of thousands of pieces of content
10x content creation scaling
Who I worked with
Engineers, sales, account management, C-suite, biz-dev, marketing
What I’d change
Focus more on how users interacted with the UI
Advocate for gathering user insights and make data-informed content decisions
Metadata Taxonomy
problem
Metadata was inconsistent, had overlap, misspellings, and was sometimes way too specific to be useful.
solution
Our developers exported a list of our entire tag database, and then I worked with another member of the content team to remove and consolidate hundreds of tags and group them into categories. Then, i grouped those categories to be consistent with our user interface.
Result
A more consistent list of metadata, which was eventually used to help a content ingestion/scaling tool.
Who I worked with
Developers, other members of the content team
What I’d change
I got a little in the weeds putting the tags into the intermediate categories—all in the name of a project that never ended up happening. So instead, I’d focus more on the consolidation and general categories.