An original, chapter-by-chapter companion to Speech and Language Processing by Daniel Jurafsky and James H. Martin, 3rd edition draft. Written as a first-principles story for master's students: every finished chapter is at least 3,000 words with the conceptual map, derivation, worked case, evaluation design, failure modes, study lab, oral-exam questions, and a complete summary.
The map of language technology: what NLP and speech systems do, why language is difficult, and how the book connects models, data, linguistic structure, and evaluation.
chapter 2Words, morphemes, Unicode, subword tokenization, corpora, regular expressions, rule-based tokenization, and minimum edit distance.
chapter 3Conditional probability, n-grams, training and test splits, perplexity, sampling, overfitting, smoothing, interpolation, backoff, and entropy.
chapter 4Supervised classification, the sigmoid and softmax, features, cross-entropy, gradient descent, precision, recall, F1, cross-validation, and significance testing.
chapter 5Lexical semantics, distributional meaning, count vectors, cosine similarity, word2vec, semantic properties, bias, and evaluation.
chapter 6Units, nonlinear activation, XOR, feedforward networks, classification, embedding inputs, backpropagation, optimization, and regularization.
chapter 7Language-model architectures, conditional generation, prompting, decoding, pretraining, scaling, evaluation, ethical risk, and safety.
chapter 8Self-attention, transformer blocks, parallel computation, token and positional embeddings, language-model heads, sampling, training, scaling, and interpretation.
chapter 9Supervised instruction tuning, preference data, reward modeling, RLHF-style optimization, direct preference optimization, alignment limits, reasoning, and test-time compute.
chapter 10Bidirectional transformer encoders, masked-token training, contextual embeddings, fine-tuning for classification, and sequence labeling.
chapter 11Sparse and dense retrieval, inverted indexes, ranking, evaluation, question answering, retrieval-augmented generation, grounding, and datasets.
chapter 12Cross-language divergence, encoder-decoder translation, attention, beam search, low-resource translation, evaluation, and bias.
chapter 13Recurrent neural networks, sequence modeling, stacked and bidirectional architectures, long short-term memory, encoder-decoder models, and attention.
chapter 14Speech sounds, phonetic transcription, articulation, prosody, acoustic signals, spectrograms, log-Mel features, and MFCCs.
chapter 15The ASR task, convolutional encoders, encoder-decoder recognition, self-supervised speech models, CTC, decoding, and word error rate.
chapter 16TTS pipelines, audio codecs and discrete tokens, language-model speech generation, VALL-E-style two-stage systems, evaluation, other speech tasks, and spoken language models.
Word classes, part-of-speech tagging, named entities, HMMs, conditional random fields, neural sequence labeling, and span-level evaluation.
chapter 18Constituents, context-free grammars, treebanks, ambiguity, normal forms, CKY dynamic programming, neural span parsing, and evaluation.
chapter 19Head-dependent relations, dependency trees, transition-based parsing, graph-based parsing, projectivity, decoding, and attachment evaluation.
chapter 20Relation extraction, event extraction, temporal representation, aspect, TimeBank-style annotation, temporal analysis, and template filling.
chapter 21Semantic roles, alternations, thematic-role limits, PropBank, FrameNet, SRL systems, selectional restrictions, and predicate decomposition.
chapter 22Emotion definitions, sentiment and affect lexicons, human labeling, semi-supervised induction, supervised word sentiment, lexicon-based recognition, entity-centered affect, and connotation frames.
chapter 23Mentions, anaphora, coreference clusters, mention detection, ranking architectures, entity linking, evaluation, Winograd-style cases, and gender bias.
chapter 24Coherence relations, discourse structure, centering, entity-based coherence, local representation learning, global coherence, and argument structure.
chapter 25Turn-taking, adjacency pairs, grounding, repair, dialog acts, conversational corpora, context, initiative, and evaluation.