Learning about AI/ML for Text, with Wordle!
summary
Uses Wordle as a concrete text problem and shows working Python code that pulls five‑letter words from the Ubuntu dictionary, scores candidates by letter frequency and TF‑IDF, and finds raise/arise as strong starters. Demonstrates that a quick multinomial Naive Bayes classifier and k‑means clustering perform poorly, that GloVe/BERT embeddings capture meaning but not helpful for letter puzzles, and that a reinforcement approach via Wordle Gym learns to win reliably under six guesses though typically one turn worse than the best stat methods. Code and data live on GitHub.
AI-written summary · about
chapters (10)
- 0:00Introduction to AI and ML
- 6:31What is Wordle?
- 8:59Do you need AI for Wordle?
- 14:18Turning text into numbers
- 17:10Finding the best starting word
- 20:39Building a Wordle bot
- 22:49Word embeddings and semantics
- 27:44Word2Vec and ELMo explained
- 30:13Reinforcement learning for Wordle
- 31:40Summary and next steps
related talks
24mCracking Wordle: Machine Learning based Strategies
53mUsing GPT Visual Capabilities to Solve a Wordle Puzzle
19mNew York Times' Connections: A Case Study on NLP in Word Games
15mAgentic Reasonings with Reinforcement Learning
48mBuilding an AI/ML powered text search system
31mTaking AI/ML to the next level: Text!
44mTeaching Computers ABCs: A Quick Intro to Natural Language Processing
30mKick-Start your Understanding of Machine Learning with Python