Making machines understand and generate language — vector semantics and word embeddings, n-gram and neural language models, neural networks, and part-of-speech tagging from HMMs to Bi-LSTM-CRF and transformer/LLM approaches. This vault holds the same resources as every subject: cheatsheet, companion notes, question bank, formula sheet, book explained, and references. Covers Sessions 1–7 (mid-sem).
The full mid-sem paper worked end-to-end — collapsible Q&A — plus a makeup-exam study guide on the same concepts.
01 · cheatsheetDense one-glance reference — TF-IDF, PPMI, Word2Vec, n-gram LMs, smoothing, neural units, HMM/Viterbi, Bi-LSTM-CRF. Sessions 1–7.
02 · companion notesA flowing, first-principles article on how machines represent and model language — embeddings, language models, neural nets, tagging. Sessions 1–7.
03 · question bankExam-style questions grouped by topic. Questions only — no answers — plus worked numerical exercises. Sessions 1–7.
04 · formula sheetEvery equation for the mid-sem — cosine, TF-IDF, PPMI, softmax/skip-gram, chain rule, n-gram MLE, smoothing, perplexity, forward/Viterbi — symbols defined.
05 · book explainedJurafsky & Martin (SLP3) — Ch 3, 6, 7, 8 — chapter by chapter, in easy language with diagrams, worked examples and gotchas.
06 · referencesTextbook (SLP3), Stanford CS224n, and the key papers — Word2Vec, GloVe, Viterbi/HMM, Bi-LSTM-CRF, BERT — worth keeping.
01 · CHEATSHEET
A dense, scannable card: distributional semantics, term-document and word-context matrices, TF-IDF and PPMI, cosine similarity, Word2Vec (skip-gram/CBOW, negative sampling), GloVe, n-gram language models and smoothing, perplexity, neural units and feedforward nets, HMM POS tagging and Viterbi, Bi-LSTM-CRF — to cram from before the mid-sem.
02 · COMPANION NOTES
A long-form, first-principles article — not a slide recap — that teaches the concepts directly: how meaning becomes a vector, why embeddings generalize, how language models assign probabilities, what neurons compute, and how part-of-speech tagging works from HMMs to neural sequence models, with the why spelled out.
03 · QUESTION BANK
Questions covering all of Sessions 1–7 — questions only, no answers — grouped by topic, for active recall and exam practice. Work them cold, then check yourself against the notes.
04 · BOOK EXPLAINED
Companion notes for Jurafsky & Martin — Speech and Language Processing (3rd ed.) — N-gram LMs (Ch3), Vector Semantics & Embeddings (Ch6), Neural Networks (Ch7) and Sequence Labeling / POS Tagging (Ch8) — chapter by chapter, in plain language with diagrams and worked examples.
05 · REFERENCES
Textbook (Jurafsky & Martin SLP3), the foundational papers (Mikolov's Word2Vec, Pennington's GloVe, Viterbi, Huang's Bi-LSTM-CRF, BERT), lecture series (Stanford CS224n), and extra reads worth keeping.