Beyond Bag of Words A Practitioner’s Guide to Advanced NLP
summary
Builds a movie recommender and a time‑series anomaly detector with runnable Python code. Uses the MovieLens 10M dataset, scikit‑learn pipelines and K‑Nearest‑Neighbors (metric=cosine) to compare content (genres + 20 neighbors) and collaborative approaches, one‑hot encodes tags into ~16,000 tag features reduced to ~120 via SVD, applies Bayesian smoothing and regularization, and reports about a 20% RMSE improvement over a mean baseline. Code and notebooks live on the conference GitHub repo for reproduction.
AI-written summary · about
chapters (13)
- 0:00Introduction
- 2:09Project overview
- 3:18Content-based recommendations
- 7:16Nearest neighbors algorithm
- 10:27Testing nearest neighbors
- 12:35Adding user tags
- 16:26High-dimensional tag challenges
- 18:18Dimensionality reduction
- 25:39User-based filtering
- 35:21Bayesian smoothing
- 38:35Rating prediction regression
- 44:11Model tuning
- 45:34Summary
related talks
01h 21mUnlocking Personalization: A Deep Dive into Modern Recommendation Algorithms
33mPicking What to Watch Next: Build a Recommendation System
57mHands-On with LLM-Powered Recommenders: Hybrid Architectures
38mBuilding Recommender System in PHP8
01h 48mRecommendation engines explained
39mHow we Used NLP and Django to Build a...
15mWhat they didn’t teach you in Grad School #bbuzz
01h 31mHands on - Build a Recommender system