Natural Language Processing

reading · we frame, you read MIT or the canonical taught · we author, no canonical fits ↺ spirals back to earlier lessons

Course locked

Complete Sequence Models and Generative Models first.

This course unlocks once you've finished its prerequisite. Open prerequisite →

Prerequisites

Sequence Models and Generative Models not yet

Complete prerequisites first to enroll.

reading · 35 min + ~212 min canonical

Text as Data: Tokenization and the Messiness of Language

Every machine-learning model you've built so far took numbers as input — pixels, feature vectors, measurements. But language arrives as text: a stream of characters with no inhere…

primary Stanford CS224N NLP with Deep Learning (YouTube)

preview →

reading · 40 min + ~212 min canonical L1

Word Embeddings: Word2Vec, GloVe, and Meaning as Geometry

You have tokens now (last lesson) — but a token like "king" is just a symbol, an arbitrary integer ID, with no more meaning to a model than "banana" or "42." To a computer, "king"…

primary Stanford CS224N NLP with Deep Learning (YouTube)

after L1 locked

reading · 35 min + ~212 min canonical L1

The Classical Toolbox: Language Models, Tagging, and Named Entities

Before the transformer, before BERT and GPT, NLP was built on a classical toolbox — and two of its tools are so foundational that modern LLMs are, in a real sense, the same idea s…

primary Stanford CS224N NLP with Deep Learning (YouTube)

after L2 locked

reading · 35 min + ~212 min canonical L1

Text Classification: The Workhorse That Ships

Of all the tasks in NLP, one ships more than any other: text classification — assign a category to a piece of text. Is this email spam or not? Is this review positive, negative, o…

primary Stanford CS224N NLP with Deep Learning (YouTube)

after L3 locked

reading · 35 min + ~212 min canonical L1

Generation Tasks: Translation, Summarization, and Question Answering

The tasks so far have mostly labeled text — classify it, tag it, extract from it. The hardest tasks in NLP generate text — produce new, fluent, correct language as output. Three o…

primary Stanford CS224N NLP with Deep Learning (YouTube)

after L4 locked

exercise reading · 15 min L1

Exercises: Easy (Warm-up)

- [ ] Implement text preprocessing pipeline in Python — Tokenization, lowercasing, stopword removal, stemming, lemmatization - [ ] Train Word2Vec from scratch on a small corpus — …

after L5 locked

exercise reading · 30 min L1

Exercises: Medium (Core)

- [ ] Fine-tune BERT for sentiment analysis — Hugging Face Transformers, IMDB dataset, evaluate accuracy/F1 - [ ] Build a Named Entity Recognition system — Fine-tune BERT for NER …

after L6 locked

exercise reading · 45 min L1

Exercises: Hard (Stretch)

- [ ] Build a question-answering system — Fine-tune BERT on SQuAD, extract answer spans from context - [ ] Implement semantic search — Sentence transformers, FAISS index, cosine s…

after L7 locked

exercise reading · 60 min L1

Mini-Project

- [ ] Build a document Q&A system in Python — Load PDFs, chunk text, embed with sentence-transformers, store in FAISS, retrieve relevant chunks, generate answers with a local LLM;…

after L8 locked

L10

project reading · 360 min L1

Project: Text Processing Pipeline

Build a Python NLP toolkit from scratch: tokenizer (word and subword), stopword removal, stemmer (Porter's algorithm), TF-IDF vectorizer, and cosine similarity search. Process a c…

after L9 locked

L11

project reading · 300 min L1

Project: Sentiment Classifier

Fine-tune BERT for sentiment analysis on IMDB reviews using Hugging Face Transformers. Implement the full pipeline: data loading, tokenization, model setup, training with Trainer …

after L10 locked

L12

project reading · 600 min L1

Capstone: AI-Powered Code Review Assistant

- [ ] Milestone 1: Build the knowledge base — ingest Ruby style guide, Rails security guide, RuboCop rules, OWASP guidelines as chunks in a vector store - [ ] Milestone 2: PR diff…

after L11 locked

L13

assessment reading · 56 min L1

Self-Assessment: Natural Language Processing

- [ ] What is TF-IDF? How does it rank documents for a search query? - [ ] Explain Word2Vec. What is the difference between CBOW and Skip-gram? - [ ] What are contextual embedding…

after L12 locked

13 lessons. Read in order; spiral back when you need to. By the end you'll have used the core ideas twice — once on the abstract, once on something you'll meet at work next week.