Speech and Language Processing (3rd Edition Draft)

Livro digital

Título:
Speech and Language Processing (3rd Edition Draft)

Autor:
Daniel Jurafsky, James H. Martin

Categoria:
Tecnologia > IA

Doador:
Raffaello D. N.

Sinopse:
This draft opens with a very specific roadmap: Large Language Models, starting with introduction, words and tokens, n-grams, logistic regression, embeddings, neural networks, then moving into transformers, masked language models, and post-training. That sequence makes the reader feel the book’s core question immediately, how do we move from basic text units to models that can generate, adapt, and reason over language at scale? The table of contents shows a deliberately broad treatment of modern NLP, not just LLMs in isolation. It covers tokenization, corpora, smoothing, evaluation, vector semantics, classification, sampling, alignment, retrieval-based models, machine translation, and the older sequence-model family, alongside speech feature extraction and automatic speech recognition. The structure also suggests a strong teaching orientation, with summaries, historical notes, and exercises supporting each chapter. What stands out most is the combination of classical foundations and current practice. This is not only a guide to transformers, but a full introduction to the linguistic and statistical machinery behind them, including parsing, information extraction, coreference, discourse, and conversation. Readers come away with a map of the field, enough technical grounding to follow contemporary systems, and a sense of how the major ideas fit together across text and speech.