Content Classification: NLP vs. AI in the Age of Data Vectorization
summary
Argues that governed taxonomies plus rule‑based NLP deliver repeatable, auditable classification that reaches mid‑90s accuracy after a human‑curated golden set, while out‑of‑the‑box ML and embeddings sit around 75% baseline. Describes tools tested (Semaphore, Hugging Face), a Slack feedback loop for human validation that rarely triggers but justifies work, dashboards for content inventory, schema.org mapping for SEO, LLMs used for headline rewrites and ADA checks, RAG/vectorization working only at very large scales (tens of thousands of documents).
AI-written summary · about
chapters (8)
- 0:00Introduction and overview
- 3:02Manual tagging and early taxonomy
- 7:24Content types and enterprise taxonomy
- 11:41NLP vs machine learning accuracy
- 20:53Popular technologies deep dive
- 27:19Taxonomy benefits and limitations
- 32:16AI's content comprehension gaps
- 33:01Q&A session
related talks
18mData Classification at Scale: Taming the Hydra
31mTaking AI/ML to the next level: Text!
46mWhy data-driven methods will shape the future (...)
08mLeveraging data insights to increase efficiency and innovation
17mNatural Language Processing in Automation
55mDemystifying Natural Language Processing across several industry verticals
35mIntelligent Ruby: Getting Started with Machine Learning
38mAutomatic DB annotation with ML tools and its problems