Book

Speech and Language Processing, Chapter 16: Automatic Speech Recognition (draft)

Daniel Jurafsky and James H. Martin, Stanford University, online manuscript of the third edition, 2026

Read the original

Read on October 17, 2026

What it is: Chapter 16 of a widely used university textbook on language processing, in its August 2026 draft. It introduces automatic speech recognition, the task of converting speech into words, and lists the practical uses that motivate it.
How much weight it carries: A standard university textbook by two long-established authors in the field, published as a draft that the authors revise. It describes the state of the field rather than measuring any product, and the wording may change in later drafts.

Facts taken from this source

The third-edition draft of Jurafsky and Martin's textbook says the input to a speech recognizer is a series of acoustic waves.
“The input to a speech recognizer is a series of acoustic waves” (Chapter 16, automatic speech recognition)
The 2026 draft of Jurafsky and Martin's textbook says that understanding spoken language, or at least transcribing it into writing, is one of the earliest goals of computer language processing.
“Understanding spoken language, or at least transcribing the words into writing, is one of the earliest goals of computer language processing.” (Chapter 16, opening paragraph)
The 2026 draft of Jurafsky and Martin's textbook says that automatic speech recognition has matured to the point where it is viable for many practical tasks, while automatic transcription by any speaker in any environment is still far from solved.
“ASR technology has matured to the point where it is now viable for many practical tasks.” (Chapter 16, opening paragraphs)
The 2026 draft of Jurafsky and Martin's textbook says speech is a natural interface for communicating with appliances, digital assistants and chatbots, especially on cellphones, where keyboards are less convenient.
“Speech is a natural interface for communicating with appliances, or with digital assistants or chatbots, especially on cellphones, where keyboards are less convenient.” (Chapter 16, opening paragraphs)
The 2026 draft of Jurafsky and Martin's textbook says transcription is important in fields like law, where dictation plays an important role.
“Transcription is important in fields like law where dictation plays an important role.” (Chapter 16, opening paragraphs)
Jurafsky and Martin's 2026 draft says hidden Markov models seem to have been applied to speech independently at two laboratories around 1972.
“Hidden Markov models seem to have been applied to speech independently at two laboratories around 1972.” (Chapter 16, history of automatic speech recognition)
The HMM, with Gaussian mixture models as the phonetic component, spread through the speech community and became the dominant paradigm by the 1990s.
“The use of the HMM, with Gaussian Mixture Models (GMMs) as the phonetic component, slowly spread through the speech community, becoming the dominant paradigm by the 1990s.” (Chapter 16, history of automatic speech recognition)
In a hybrid architecture, a feedforward neural network's outputs are used as the probability estimates for an HMM-based architecture.
“whose outputs are used as probability estimates for an HMM-based architecture” (Chapter 16, history of automatic speech recognition)
By 2012, the performance of hybrid systems, which pair neural networks with HMMs, had surpassed traditional HMM/GMM systems.
“by 2012, the performance of hybrid systems had surpassed traditional HMM/GMM systems” (Chapter 16, history of automatic speech recognition)
Jurafsky and Martin's draft says encoder-decoder speech models implicitly learn a language model from their training data.
“encoder-decoders implicitly learn a language model for the output domain of letters from their training data” (Chapter 16, automatic speech recognition)
Word error rate can be greater than 100 percent, because its equation counts insertions as errors.
“note that because the equation includes insertions, the error rate can be greater than 100%” (Chapter 16, automatic speech recognition)

Pages that rest on this source

Book

Speech and Language Processing, Chapter 16: Automatic Speech Recognition (draft)

Daniel Jurafsky and James H. Martin, Stanford University, online manuscript of the third edition, 2026

Read the original

Read on October 17, 2026

What it is: Chapter 16 of a widely used university textbook on language processing, in its August 2026 draft. It introduces automatic speech recognition, the task of converting speech into words, and lists the practical uses that motivate it.
How much weight it carries: A standard university textbook by two long-established authors in the field, published as a draft that the authors revise. It describes the state of the field rather than measuring any product, and the wording may change in later drafts.

Facts taken from this source

The third-edition draft of Jurafsky and Martin's textbook says the input to a speech recognizer is a series of acoustic waves.
“The input to a speech recognizer is a series of acoustic waves” (Chapter 16, automatic speech recognition)
The 2026 draft of Jurafsky and Martin's textbook says that understanding spoken language, or at least transcribing it into writing, is one of the earliest goals of computer language processing.
“Understanding spoken language, or at least transcribing the words into writing, is one of the earliest goals of computer language processing.” (Chapter 16, opening paragraph)
The 2026 draft of Jurafsky and Martin's textbook says that automatic speech recognition has matured to the point where it is viable for many practical tasks, while automatic transcription by any speaker in any environment is still far from solved.
“ASR technology has matured to the point where it is now viable for many practical tasks.” (Chapter 16, opening paragraphs)
The 2026 draft of Jurafsky and Martin's textbook says speech is a natural interface for communicating with appliances, digital assistants and chatbots, especially on cellphones, where keyboards are less convenient.
“Speech is a natural interface for communicating with appliances, or with digital assistants or chatbots, especially on cellphones, where keyboards are less convenient.” (Chapter 16, opening paragraphs)
The 2026 draft of Jurafsky and Martin's textbook says transcription is important in fields like law, where dictation plays an important role.
“Transcription is important in fields like law where dictation plays an important role.” (Chapter 16, opening paragraphs)
Jurafsky and Martin's 2026 draft says hidden Markov models seem to have been applied to speech independently at two laboratories around 1972.
“Hidden Markov models seem to have been applied to speech independently at two laboratories around 1972.” (Chapter 16, history of automatic speech recognition)
The HMM, with Gaussian mixture models as the phonetic component, spread through the speech community and became the dominant paradigm by the 1990s.
“The use of the HMM, with Gaussian Mixture Models (GMMs) as the phonetic component, slowly spread through the speech community, becoming the dominant paradigm by the 1990s.” (Chapter 16, history of automatic speech recognition)
In a hybrid architecture, a feedforward neural network's outputs are used as the probability estimates for an HMM-based architecture.
“whose outputs are used as probability estimates for an HMM-based architecture” (Chapter 16, history of automatic speech recognition)
By 2012, the performance of hybrid systems, which pair neural networks with HMMs, had surpassed traditional HMM/GMM systems.
“by 2012, the performance of hybrid systems had surpassed traditional HMM/GMM systems” (Chapter 16, history of automatic speech recognition)
Jurafsky and Martin's draft says encoder-decoder speech models implicitly learn a language model from their training data.
“encoder-decoders implicitly learn a language model for the output domain of letters from their training data” (Chapter 16, automatic speech recognition)
Word error rate can be greater than 100 percent, because its equation counts insertions as errors.
“note that because the equation includes insertions, the error rate can be greater than 100%” (Chapter 16, automatic speech recognition)