Partner with Econify to strategically use full stack, AI, mobile, OTT, and cloud engineering to solve complex challenges
Challenge
For media companies, article taxonomies aren’t just internal tools—they’re critical infrastructure. From powering on-site search and ad targeting to fueling BI insights and marketing personalization, structured classification is fundamental to revenue.
Yet, malformed, ambiguous, or missing taxonomy tags can be alarmingly common:
Malformed: “Jim Cramerr” vs. “Jim Cramer”
Ambiguous: “Jim” vs. “Jim Cramer”
Missing: No tag at all for “Jim Cramer”
These inconsistencies ripple downstream, undermining:
Ad targeting systems that rely on taxonomy for relevance
Search engines that surface inaccurate or irrelevant content
Marketing campaigns that mistarget users
BI dashboards that make flawed strategic recommendations
Manual editorial tagging is time-consuming, error-prone, and often skipped—making it a perfect candidate for automation.
Goals
Build a fast, low-friction proof of concept for NBC that demonstrated AI could automate taxonomy generation, improving consistency and reducing editorial workload.
This was intended as an internal demo: minimal setup, practical output, and ready for hands-on exploration by stakeholders.
Solution
Econify built a streamlined internal tool with:
- A lightweight UI
- A backend service that exposed a single API endpoint: /generate — an LLM-wrapped service call via AWS Bedrock
The service ingests unstructured article content and returns clean, consistent taxonomy suggestions—automatically resolving ambiguity and typos.
To support internal scaling and future reuse, we also built and open-sourced generative-ts—a Typescript library that simplifies working with generative models in predictable, repeatable ways.
